Time  Nick         Message
00:02 jas_         !tell nerzhul network_asio works OK on LAN, can connect to server on remote host A-OK.  when I connect to unix-experience.fr, though, i get "connection lost" dialog.
00:02 ShadowBot    jas_: O.K.
00:22 jas_         https://github.com/jastevenson303/dcbl_csm/blob/master/init.lua#L31 <-- it's funny, because yaw keeps increasing until you turn the mouse the other way :O (didn't expect that hehe) gn o/
01:48 Aerozoic     Anybody know what causes this server crash error? ERROR[Main]: ServerError: AsyncErr: environment_Step: Runtime error from mod 'default' in callback environment_Step(): /usr/local/share/minetest/builtin/common/vector.lua:16: attempt to index local 'a' (a nil value)
10:13 IhrFussel    Am I understanding this PR correctly and sofar assumes that most servers have a MAX LAG value of 0.2??? https://github.com/minetest/minetest/pull/6504
10:13 IhrFussel    That is completely and utterly WRONG
10:14 IhrFussel    Even Xanadu has a max_lag of 0.9 sometimes with 10+ players tenplus1 said recently
10:15 IhrFussel    To use the max_lag as "goodness" indicator is a bad idea in general IMO ... you penalize players with lots of mods + players that way, not sure if that's what you want to achive
10:15 IhrFussel    servers with lots of mods/players*
10:20 IhrFussel    How abou adding a new average lag to the engine? I'm pretty sure that is how most other games do it on servers
10:22 IhrFussel    If I look at the lag values my own mod tells me then it makes far more sense: ticks are set to 0.2 secs so 0.2 average means pretty much perfect... now sometimes it goes up to 0.25 or even 0.3 average which then indicates quite a lot of "hangs"
10:23 Krock        max_lag is kind of an average value already
10:23 Krock        average of maximal lag, however.
10:23 IhrFussel    Krock, it is too slow, it does not give the player a "now" value
10:24 IhrFussel    If the max_lag spikes to 30 secs (cause of WE) it takes up to TWENTY minutes to go back to normal again
10:24 Krock        average values are never a "now" value
10:24 IhrFussel    The max_lag only goes down by 0.000099 or something per tick btw
10:24 IhrFussel    It's not an average
10:25 Krock        it should decrease by half in 5 minutes
10:25 Krock        so 1/4 in 10 minutes and 1/8 in 20 minutes
10:25 Krock        https://github.com/minetest/minetest/blob/master/src/server.cpp#L491
10:25 IhrFussel    My mod still provides more useful lag values of the recent minute that's why I advise my players to use the own lag command
10:26 IhrFussel    And my problem is the PR if it gets merged, like I said servers with higher load will ALWAYS have higher max_lag
10:27 IhrFussel    The 90+ servers with 0.15 max_lag are probably DEAD ones
10:27 IhrFussel    My server is mostly pendling between 1.2 and 2.5 secs max_lag which is alright for such a loaded server
10:29 IhrFussel    At least no player complained about lag yet so I guess it itsn't major
10:36 IhrFussel    The max_lag can also spike depending on what players are doing on the map...it is a not more fair compared to ping..it is even MORE unpredictable
10:39 IhrFussel    Example: max_lag was 3.2 but my lag mod showed only ONE TICK took longer than 0.5s and ALL other ticks were perfect ... now players still see 3.2 in /status and (if the PR gets merged) in the server list
10:39 Krock        so renaming max_lag to lag and simply calculating the average would be more suitable?
10:39 Krock        which PR are you talking about btw?
10:40 IhrFussel    Krock, https://github.com/minetest/minetest/pull/6504
10:41 IhrFussel    And yes if the lag value in /status was more accurate for the recent minutes I'd be fine but not if it takes 10+ minutes to go back to normal after 1 large spike
10:42 Krock        that PR looks fine. The lag calculation behind just needs tewaking
10:43 IhrFussel    Also I find the lag values in his PR to be completely nonsense...like I said even xanadu has a lag of 0.9 sometimes tenplus1 said so if you really merge that PR you'd need to adjust the good/medium/bad indictaors accordingly
10:43 IhrFussel    You cannot expect from a server to ALWAYS run at the set tick rate
10:44 IhrFussel    That's not how it works on a server with many mods
10:44 Krock        some spikes are guaranteed as soon you have moretrees or another mod that calculates the terrain height or places large structures
10:45 IhrFussel    Yes and those servers now will get bad indictaors just because of those mods?
10:46 IhrFussel    That's why I think the max_lag value needs to be more accurate to the current server situation...that way lag spikes caused by mods aren't present after a few minutes anymore
10:47 IhrFussel    And thanks by the way for the unknown node pos PR :)
10:54 shivajiva    IhrFussel: looks like a QOS indicator and will not stop players joining to see their friends, might get the dead servers the odd player though
10:54 Fixer        today i've learned minetest has SILLY ping indicator that bullshits everyone not living on serverlist continent :/
10:56 IhrFussel    shivajiva, I simply think it is kind of discriminating to give servers with 0 or 1 player and maybe 10 mods a good indicator and servers that built-up over time and therefore have quite a few mods and maybe also lots of players a bad one
10:58 IhrFussel    The internal server lag doesn't matter the same on each server...creative servers largely don't care about lag I'd guess cause they need no "realtime" reaction
10:58 shivajiva    you will get used to the bullshit ways some things are handled, in my experience none of it matters to the players, only the quality of their gameplay and playing in a dead server doesn't preclude anything other than loneliness ;)
11:01 shivajiva    as I recall skyblocks and a couple of other servers spent so much time at the top of the list that they decided it wasn't fair and re-weighted it
11:02 shivajiva    even though it's status at the top reflected it's popularity
11:04 shivajiva    the list is just being sorted in diff ways to please diff people, clearly it offends sofar, wonder if his server is hosted in the US
11:06 shivajiva    I'd like to see the ping and lag displayed in game in real time personally between the client and server
11:10 IhrFussel    shivajiva, but do you agree that the max_lag value should be a bit more accurate and not display numbers that happened 5+ minutes ago?
11:11 shivajiva    we need some latency or the stability of the overall picture is distorted equally
11:11 IhrFussel    Or we could just add a new avg_lag value to /status so we have both avg and max over time
11:13 IhrFussel    My server utilizes the areas mod and WHENEVER a player modifies a protection the server lags for 1+ sec CAUSE of the large protection file ... that is not my fault and I refuse to accept it as a fault of my server
11:14 IhrFussel    I could name many more examples where I refuse to fault my code for it
11:14 shivajiva    yea a current average would be good, wonder if I should investigate db storage for areas
11:14 IhrFussel    to blame*
11:16 shivajiva    db access becomes viable once the data set gets beyond a certain size and areas can become extremely big
11:16 shivajiva    I can't see the first 400 records in the list for skyblocks it's become so big
11:16 Calinou      hi
11:16 shivajiva    Hi Cal o/
11:17 Calinou      by the way, how many of the servers use PostgreSQL?
11:17 Calinou      I've been using it and it works well so far
11:17 IhrFussel    I already cleaned up my areas.dat file...I removed any area owned by a account that got deleted...but it's still ~ 7,000
11:18 shivajiva    there would be a significant gain from holding the data in a way that can be accessed quickly but adds little overhead to minetest
11:19 shivajiva    Cal: only one I haven't tried
11:19 Calinou      I'd argue most servers should use PostgreSQL
11:19 Calinou      SQLite has its limits, and using it on servers is usually a bad idea
11:19 shivajiva    go on then :)
11:20 Calinou      (because you might as well use RDBMSes)
11:20 Calinou      client-server ones, that is
11:22 IhrFussel    Anot example: With ~ 60,000 accounts in auth.txt WHENEVER a player joins the engine updates the last login timestamp so it writes the auth file anew AFAIK ... and that lags my server ~ 0.2 sec
11:22 IhrFussel    Another*
11:23 shivajiva    give sauth a try fussel
11:23 Calinou      yeah, prune unneeded entries
11:24 IhrFussel    Is that a mod or what is it?
11:25 Calinou      by the way, https://www.phoronix.com/scan.php?page=news_item&px=PostgreSQL-10.0-Released:)
11:25 Calinou      just noticed
11:25 shivajiva    https://github.com/shivajiva101/sauth
11:34 Shara        Hello all
11:34 Krock        hi Shara
11:34 Shara        Hi Krock :)
11:35 Shara        shivajiva: areas on RC scare me as you know. I somehow feel like that will break one day.
11:35 Krock        Calinou, so we're doing something wrong? SQLite is the default backend for ages
11:35 Calinou      yes, in fact it was wrong since the beginning (or almost) :P
11:35 Calinou      at least, recommending it for use on servers is a bad idea
11:35 Calinou      it's fine for singleplayer
11:36 Calinou      setting up PostgreSQL is not that hard
11:36 Calinou      https://wiki.minetest.net/Database_backends
11:36 Calinou      I should write some docs on setting up PostgreSQL for Minetest
11:37 Krock        but.. dummy is fastest :P
11:38 Calinou      :D
11:38 Calinou      I used that one on my CTF server, since world data didn't have to be stored
11:39 Krock        indeed, helpful there. Also for testing mapgen it's nice to tweak parameters and rejoin
11:40 Calinou      it's a shame most popular servers run on underpowered machines, causing lag
11:40 Calinou      I wish there was some kind of voting on servers, too
11:40 Calinou      (simple upvotes/downvotes, like, every player can vote once per sever)
11:40 Calinou      server*
11:40 Calinou      so you could vote based on your experience or something
11:41 Krock        which, then again, would require a master server to bundle all the information
11:41 IhrFussel    Calinou, the problem is that MT uses 1 core only mainly
11:41 Calinou      that's not the only problem, the default settings are quite bad
11:41 IhrFussel    I'm pretty sure that my server would boost its performance quite a bit if it was able to use all 4 cores
11:41 Calinou      look at https://gist.github.com/Calinou/683cb0748efe867be6e18f86f206a6b7, compare it with the default settings
11:42 Calinou      also, servers peform better on faster CPUs (single-thread performance), so a dedicated is better than a VPS, too
11:42 Calinou      (most sizeable Minecraft servers run on dedis)
11:42 Calinou      my server is a VPS, but I'd go dedi if I had 15+ players constantly
11:42 IhrFussel    Mine runs on a dedi but not on a $100 one
11:42 Calinou      there's dedi and dedi :P
11:42 Calinou      the Atom ones are pretty slow
11:42 IhrFussel    Opteron
11:43 Calinou      Opterons are bad at single-thread performance
11:44 Krock        is changing full_block_send_enable_min_time_from_building a good idea? it will increase the traffic
11:44 IhrFussel    It can handle MT pretty good but since I needed so many mods (160+) of course it struggles
11:44 Calinou      Krock: I don't think that setting is relevant since we have client-side node placement prediction
11:44 Calinou      it used to not be the case (before 0.4)
11:44 Calinou      any non-zero value will increase perceived lag, since players who need to receive MapBlocks will not have them immediately
11:45 Calinou      more blocks need to be sent after all, I increased the view/generate distances to 14 :)
11:45 Calinou      (Minecraft uses 15, so I'm still lower than that)
11:46 Krock        also, aren't minecraft blocks larger than ours? Y=256
11:46 Calinou      they're the same size, but height is only 256 nodes indeed
11:47 Krock        ah
11:47 IhrFussel    What? They still didn't increase Y?
11:48 IhrFussel    What makes them think that buildings don't need to be higher than 256? Ridiculous
11:48 Calinou      yeah, Y is 256 in Minecraft since March 2012 :P
11:48 Calinou      they tried 512, but it was too slow
11:48 Krock        width of an insane s32 is more important!!!!
11:48 Calinou      Krock: i128 or bust!
11:48 lisac        we've got a taller building on Craigs
11:48 lisac        and it's also huge
11:49 lisac        https://forum.minetest.net/viewtopic.php?p=227356#p227356
11:49 IhrFussel    Too slow? Then they're doing something wrong in the code I'd say...why would the max height of the map slow down map loading?
11:49 Krock        we can "easily" rebuild the mt. everest and Burj Khalifa
11:50 Krock        no chance for you, minecrafters
11:50 IhrFussel    Are there at least mods in MC that increase the height? Or is it hard-coded?
11:50 Krock        lisac, I hope WE was used to copy floors, otherwise that's a massive task
11:51 Calinou      there was Cubic Chunks for Minecraft Beta 1.7.3
11:51 Calinou      (before the height was 256, it was just 128 back then)
11:51 Calinou      it increased height to 65,635
11:51 lisac        Krock, 100% survival
11:51 Calinou      or 4096, I'm not sure
11:51 Krock        uh, amazing
11:52 IhrFussel    Wait...are you saying MC chunks load Y 256 at ONCE?
11:52 paramat      IhrFussel a max lag indicator that is influenced by brief lag spikes is actually reasonable because a player's experence of a server tends to depend on the occasional lag spikes more than the times of lower lag. however agree with adjusting the value->colour encoding (see comment)
11:54 paramat      MC doesn't use cubic chunks, our behaviour is identical in all axes
11:54 Krock        this means they generate much more useless data than we do
11:54 Krock        unless they use a differencial method to save it to the disk
11:55 paramat      but then they have less issues caused by 'mapchunks above not being generated yet' like shadow issues etc.
11:57 Calinou      IhrFussel: I think it's 2× 128
11:57 IhrFussel    paramat, the only problem I see in the PR is that servers likely will get penalized for having (temp) higher load caused by intensive mods running or many players actively playing
11:58 IhrFussel    I wouldn't even call 0.5 sec max_lag "high" you can barely notice 0.5 sec lag
11:59 paramat      wut
11:59 paramat      0.5s is a long time
12:00 IhrFussel    IMO the calculation for max_lag right now is not good...it should calculate a REAL average and not decrease by 0.00...99 per tick if lag is low, that is no real average
12:01 IhrFussel    average means add ALL tick times and divide by their amounts
12:01 IhrFussel    amount*
12:01 paramat      max lag indicator will simply be an indicator of the 'max lag you will experience on a server' so is fair and does not discriminate. all servers will be affected the same way, and most have intensive mods like moretrees and caverealms
12:02 paramat      from what i've read about your server red seems accurate, you admit is has long lag spikes
12:03 IhrFussel    max_lag 1.5 BUT my avg lag mod reports an AVERAGE of 0.204
12:03 IhrFussel    That is a big difference
12:04 nerzhul      hello world
12:05 ThomasMonroe hi nerzhul
12:05 Fixer        IhrFussel: yes, it loads Y=256 at once
12:05 Fixer        IhrFussel: MC chunk is 16x16x256 iirc
12:06 nerzhul      Krock, #6505 okay for a merge, if you can just fix the near condition could be nice
12:06 paramat      there is no penalization, it's an indication of 'max lag you may experience', players will soon find out that the servers with more features have higher lag so will see it as a price worth paying
12:06 IhrFussel    But it is FAR from reality of the recent minutes... here https://pastebin.com/BXQ5z1X4
12:07 IhrFussel    When my avg lag mod reports 1.0 max lag in the last minute /status may still display 2.0 or more
12:07 IhrFussel    Which is not fair
12:07 Fixer        don't forget, max_lag is not a speedometer
12:08 Fixer        it works differently
12:08 nerzhul      nice Krock thanks, merge asap when you want :)
12:08 IhrFussel    Then we need a REAL avg lag indicator like I said
12:08 Fixer        you can have huge temporarely spike to say max_lag 100, and then there will be no lag while max_lag itself slowly goes down from 100 to 0 over time
12:08 Krock        nerzhul, okay. Will most likely do it this evening
12:08 IhrFussel    I rthink that is what players really want... an average of the recent lag values
12:09 IhrFussel    Don't take the *highest* lag and call it a day, that is lazy
12:09 Fixer        max_lag can be misleading
12:09 paramat      "My server utilizes the areas mod and WHENEVER a player modifies a protection the server lags for 1+ sec CAUSE of the large protection file ... that is not my fault and I refuse to accept it as a fault of my server" of course it's your fault, if that lag spike occurs then an indication of that lag spike is fair and accurate
12:10 IhrFussel    Okay so it#s my fault when my server has that many areas alright
12:10 paramat      Ihr you write so much nonsense =)
12:11 IhrFussel    You don't seem to understand that max_lag doesn't tell the player what lag awaits them NOW
12:11 Fixer        IhrFussel: it is not possible anyway to now lag up front
12:11 IhrFussel    It only tells them what the highest lag in the recent time was
12:12 Fixer        IhrFussel: https://forum.minetest.net/viewtopic.php?f=6&t=10884#p167292
12:13 IhrFussel    "So even if max_lag is showing, say, 5.6s the current server lag may be (and probably is) a lot less." <<< and that confuses players!
12:14 Fixer        IhrFussel: yeah it does
12:14 Fixer        IhrFussel: start up WE, do smth, server halts for 10 seconds, unfreezes, reports 10 sec max_lag that then starts to decline _slowly_ (but there is no lag after that ofc)
12:14 IhrFussel    Sometimes player tell me "ADMIN LAG IS HIGH" JUST because /status told them a high number
12:15 Shara        I started to educate my players about max_lag.... now they understand what it is.
12:15 Shara        These days they mostly just complain if it's increasing over time
12:15 Fixer        IhrFussel: lag _was_ high
12:16 IhrFussel    Server owners shouldn't act as a teacher here, the server list and /status info should simply be more accurate
12:16 Fixer        _max_lag does not show current server lag_
12:16 Fixer        now thats the problem
12:17 IhrFussel    That the server lagged for 10 secs cause of WE is not relevant to the player who just joined and it's even less relevant if that WE action was a rare one
12:17 paramat      as i wrote, the experience of a server tends to be determined more by the lag spikes more than the times between, so the lag spikes having a big influence on the indicator is ok, it shouldn't be an average as that would rarely ever indicate the lag spikes
12:18 Shara        I find max_lag helpful, since it tells me when there has been something unusual I should check on. Which isn't to say another way of measuring it would not be helpful as well
12:18 Shara        Stuff in f5 isn't really for a player who just joined anyway
12:18 IhrFussel    paramat, just add another value then... just make it so /status shows avg_lag=0.3, max_lag=1.0 for example
12:19 Fixer        +1
12:19 Fixer        Shara: can you give me that script that prints max_lag?
12:19 Fixer        IhrFussel: ^
12:21 IhrFussel    Fixer, http://ihrfussels-server.tk/lagmod.tar.gz ... it tells you the avg/max lag per minute but you can of course adjust the time between measurements
12:21 Fixer        nice, ty, i have an idea
12:21 IhrFussel    Also remove the depends.txt
12:21 Fixer        to visualise the problem
12:31 IhrFussel    The best way to visualize it is using WE for a big project and then comparing the minutes after to see how "fast" the max_lag goes down to the ACTUAL lag
12:31 IhrFussel    In the meantime my avg lag mod will already display the real recent lag and also the real max value from that time
12:32 Fixer        IhrFussel: yeah
12:33 Fixer        IhrFussel: that what i wanted, kinda, just to visualise the problem
12:34 Fixer        print(minetest.get_server_status()) will probably help
12:34 IhrFussel    max_lag is flawed in general because NOBODY really knows where the number started...one player joins and sees it as 20 for example...another sees it as 15 the third as 12 etc etc
12:34 Shara        IhrFussel: You actually get such high numbers?
12:35 Shara        For me, anything 10+ is a sign somerthing happened that should not have
12:35 IhrFussel    No that was an example with WE ... when you modify 10 million nodes any machine will lag 5+ secs
12:35 paramat      any single value or icon would be flawed
12:35 Shara        People shouldn't really be editing insanely large areas with WE at once on servers.
12:36 Fixer        Shara: with GC64 or lua you can
12:36 Shara        I expect anyone I give the priv to on my servers to be mindful about what will trigger lag
12:36 Fixer        GC64* is scary one though, but it works on windows
12:36 IhrFussel    Shara, sometimes you have to...but WE isn't the only concern...MANY mods sometimes cause a high lag spike and max_lag just does *not* tell the player what to expect NOW
12:38 Shara        Again, mods doing crazy things that generate high lag with any frequency, probably aren't great to have on a server to begin with
12:38 IhrFussel    I don't see why anybody could be against a second avg_lag value in /status ... it doesn't hurt anybody
12:38 Shara        I haven't said I'm against it
12:38 Shara        Just surprised it's such a big issue
12:39 Shara        It only seems to be the more technically aware players that press f5 and check these things, at least from my experience
12:39 IhrFussel    Shara, when my avg lag mod tells me < 0.3 secs average most of the time while max_lag is somewhere in between 2-3 then YES I see it as an issue
12:40 IhrFussel    Because all that says is there WAS a high lag of 2-3 seconds but nobody knows exactly WHEN or if it will happen again
12:40 Shara        I can't check my averages, since right now I have a dead auth file. Auth files corrupting feels a bit more important given the security implications of when the admin account password is no longer recorded
12:41 Fixer        i want better lag display too
12:41 Shara        Those are the kinds of issues I'd much rather see some time devoted to
12:42 IhrFussel    And if the devs won't implement a good avg_lag value I'll have to replace the ENTIRE /status output with my own
12:42 Shara        IhrFussel: why not make a PR to add something then?
12:42 Shara        That's how I got things I really cared about added
12:43 IhrFussel    It's not about creating a PR, I'm sure that can be done in a matter of 10 minutes...I simply want agreement among devs
12:43 Shara        YEa, the console option I wanted could have been done by most devs in that time too.. but two versions of me asking and no one wrote it, until I went ahead myself
12:43 IhrFussel    Cause since you might know if devs are against a PR they won't merge it => wasted time
12:43 jas_         is there already an issue on tracker for this, IhrFussel?
12:44 IhrFussel    jas_, I didn't check yet but there are surely "bug reports" in the forum about it
12:44 Shara        The devs will work on what they feel is most important, but every one of us can contribute.
12:45 jas_         but if there's no issue on tracker, then that's a good place to start, i do believe
12:45 Shara        If you can do that and just want the devs to confirm if they'd support the idea, you could ask them that directly.
12:45 IhrFussel    The max_lag issue wasn't a problem to me by the way UNTIL NOW ... cause now they want to use this misleading value for the server list
12:45 jas_         they do?
12:46 IhrFussel    Yes check the newest PR by sofar
12:46 jas_         has it been merged?
12:46 IhrFussel    Not yet
12:46 Shara        Oh, he wants his server higher I guess
12:46 jas_         PR is pull _request_
12:46 Shara        IhrFussel: link please?
12:46 IhrFussel    jas_, but 2 devs already like it
12:47 jas_         this would be a lot less of an issue, if players were able to sort columns.  is this even possible in formspec tables?  i don't think so
12:47 jas_         (maybe, never tried.)
12:47 IhrFussel    https://github.com/minetest/minetest/pull/6504
12:47 jas_         (sort by ping, sort by players, etc)
12:47 Shara        thanks
12:47 jas_         or even a "Sort" button, to select how to sort.
12:48 IhrFussel    jas_, but to use a value that only tells the players "there was some high lag in the past but nobody knows how high it is now" is a bad idea to me
12:48 jas_         : )
12:48 jas_         i know
12:49 jas_         if it's any consolation, no one is using minetest-dev on your server.  will this be merged into 0.4.17?
12:49 IhrFussel    Especially when you look at the time values... < 0.2 = good?? WHAT?
12:49 jas_         0.02 is way better
12:50 IhrFussel    The only servers that have a max_lag of <0.2 are ones with either VERY simple mods or with VERY few players
12:52 Fixer        IhrFussel: here is typical example of max_lag reading after game start: https://i.imgur.com/Z1rdY50.png
12:52 Shara        Dark Lands flags as orange with only six players there, but I know it's quite playable in general
12:53 Shara        This basically just means if you use anything other than a really powerful server or have any players... you're in trouble
12:53 IhrFussel    4.2 max_lag, 0.205 avg lag
12:53 IhrFussel    Great now any new player will think the lag is 4.2 secs
12:53 Fixer        IhrFussel: have you created issue on github?
12:54 Fixer        this may be a good start
12:54 Shara        I don't have any issues with max_lag, but that PR is kinda... ugh
12:54 jas_         (for the avg_lag feature request)
12:55 IhrFussel    Shara, does your server never go above 0.5 max_lag?
12:55 Shara        IhrFussel: I do not mind that it does
12:55 Shara        It is always above 1
12:55 Shara        (or maybe not.. I need to check again)
12:55 IhrFussel    Well you likely would mind if that meant "bad" in the server list
12:55 Shara        Did you read what I just said?
12:56 Shara        (1:54:19 PM) Shara: I don't have any issues with max_lag, but that PR is kinda... ugh
12:56 Shara        I don't like the PR. The PR is the issue, not max lag itself
12:56 IhrFussel    Yeah sorry I read it too late
12:56 Fixer        i want proper network latency display in server list
12:56 Shara        I find max_lag a useful thing to tell me when there's been a problem
12:57 Shara        Since if I see max lag is high, there;s probably something worth following up on
12:57 Shara        Heh, DL is way too high for example, yet players don't really seem to care much
12:57 jas_         haha just noticed your server in that picture on the PR Ihr, sorry
12:57 IhrFussel    It is good for server owners to see when the server hung for lots of secs i agree...but it is not useful to the player who just wants to check how reliable the gameplay/connection would be
12:58 Shara        This isn't about whether it's useful for the player. It's not for the player
12:58 Fixer        max lag calc: https://pastebin.com/raw/0Cir3QxT
12:58 IhrFussel    But players USE it as a lag indicator cause there is nothing else to use
12:59 paramat      ok i added a comment to the PR
12:59 Shara        I really don't see many, other than those quickly capable of learning how to really use it, commenting on it
13:00 paramat      also i commented that 'red' shows up too easily
13:01 Shara        sofar says busy servers should peak at 0.02 max_lag? Am I understanding correctly?
13:01 Shara        0.2*
13:01 IhrFussel    If it's not for player we need to implement a value for players...I think I will really create an issue for that later
13:01 Shara        Up to 2 on max_lag is certainly fine for playing
13:02 Fixer        what is "SEnv step" ?
13:02 jas_         "Busy servers will peak out above 0.2" <-- i believe that is a very large range, indeed
13:02 IhrFussel    Shara, paramat says 0.5 is high
13:02 Fixer        is that what we need? server step?
13:02 IhrFussel    0.5 AVERAGE would be high
13:02 Shara        You know, a lot of my players on RC tell me they love the server due to how well it runs.
13:03 Shara        I could understand DL not getting a green rating, but if RC doesn't there's something off
13:03 IhrFussel    But we aren't talking about average... "max_lag *= 0.9998; // Decrease slowly (about half per 5 minutes)" << this has NOTHING to do with average calculation
13:03 IhrFussel    At least I learned it differently in school
13:05 paramat      yes 0.2 max lag is low and doesn't deserve a red
13:05 IhrFussel    IMO max_lag < 0.5 is PERFECT, 0-5 - 1.0 is GOOD, 1.1 - 3.0 is OKAY and everything above can be "bad"
13:05 Shara        paramat: even 2 does not deserve a red, probably not even 3
13:06 Shara        5 + maybe
13:06 jas_         there's always some coloring going on in the server list now (pictures in PR screenshot).  i think sorting ability is more valuable
13:06 jas_         sort by players, alphabet, actual ping, max_lag AVERAGE
13:06 Shara        jas_: Yes! Sorting ability would be lovely
13:06 jas_         and "default" weighting
13:06 IhrFussel    You think wrong about max_lag paramat you think too much "average"
13:06 paramat      i didn't write 0.5 was high, i wrote 0.5 is easily caused by various mods
13:07 Shara        paramat: May I please recommend that once I have RC up again, you spend some time idle there just to track max lag compared to number of players
13:07 Shara        you'll see max_lag gets higher than this with the game still fully playable
13:08 jas_         what amount of time does max_lag average over, anyway?
13:08 jas_         is it ten minutes?
13:08 Shara        My servers are successful and players like them. Why should I be penalised when players don't have complaints?
13:09 IhrFussel    jas_, "max_lag *= 0.9998; // Decrease slowly (about half per 5 minutes)"
13:09 jas_         sofar's PR is a good concept, poor implementation, imho
13:09 jas_         i don
13:10 Shara        Yes, agreed
13:10 jas_         i don't know what that means, actually HHAA
13:10 jas_         but yeah, the client needs to ping all the servers, like in quake (sorry, is that right?)
13:10 jas_         to get actual ping, and then add sort buttons.  visually seeing which servers run "hot" might be nice, but there's always coloring in server list
13:10 IhrFussel    jas_, it means the value will always decrease by the same amount and if there is no more higher max_lag it would be 0 after 10 minutes
13:11 paramat      Shara "you'll see max_lag gets higher than this" higher than what value? and note i do agree with you
13:11 Shara        easily higher than 2 at peaks
13:11 Shara        Though I think much above that is rare
13:11 jas_         allowing the client to sort this list, and giving them actual ping information, seem reasonable -- why is this not so?  is it because pinging all the servers is not ideal?  in quake, you had to hit the "refresh server list" to get new ping values
13:11 jas_         (bbfn, gl hf)
13:11 Shara        Unless we're doing anything big with WE (and then we warn players in advance)
13:12 Fixer        paramat: people need display of current lag, not some smoothed smth
13:12 IhrFussel    On my high loaded server on a dedi I get max_lag between 1.5 and 2.5 mostly...depending on what players do it can peak to 4 or 5 once but then goes down to 1-2 again but it is pretty constant at 1-2 secs most of the time
13:13 Shara        It's making me want to block these app users more and more :(
13:13 IhrFussel    Shara, I sometimes get the feeling those 3rd-party-apps demand more server performance (send more data or more requests maybe)
13:14 Shara        way back, the thing that made lag felt on RC was player joins...switched to remote media and it helped
13:14 Shara        But all thos eocnstant joins are not helpful
13:15 paramat      and i agree that 'max lag' is perhaps not the most useful value to show to players on the serverlist, but ping is meaningless of course
13:15 Fixer        why can't minetest client ping itself?
13:16 Shara        Fixer: Yea, was trying to think that through
13:16 Fixer        like other games
13:16 Shara        What matters is ping for the player
13:16 paramat      Shara i don't need to join RC to experience that, i take your word for it
13:16 Shara        But that also doesn't help on server list, since everyone sees the same server list at the moment
13:16 Shara        paramat: thank you
13:16 Fixer        for example minetest can get server list and use its ping values at first but then it starts pinging himself and reordering everything, serverlist ping is backup
13:17 Shara        I assume client list syncs with main public list currently
13:17 paramat      2s of lag is annoying when it happens but is ok if it is rare, it doesn't indicate the playability of the server
13:18 Shara        paramat: exactly. But due to how max lag works, it stays at the higher value for  along time
13:18 IhrFussel    If it would I would probably have no players anymore...2 secs lag is really not *that* rare on my server but the players seem to be able to handle it just fine
13:18 Shara        I really do like and find max_lag useful as an indicator of problems, but it's not really accurate for saying how good a server is to play on
13:21 IhrFussel    max_lag can only be constantly low with mods+many players on extremely powerful servers really
13:22 IhrFussel    I already pay $25 monthly for my dedi ... and it would be wrong if another person who plays $100 or more gets a better ranking cause they paid more money
13:22 IhrFussel    pays*
13:23 Shara        Yes
13:24 IhrFussel    I already find it stupid that the amount of lag decides whether or not players can glitch through nodes on the map but nobody cared
13:24 paramat      so IhrFussel thinking on this i agree with you in a way. if players understood how max_lag works there wouldn't be much of an issue, but they mostly won't and are likely to see the icon as a playability indicator
13:25 Shara        I agree with the PR in that ping is currently not a good indicator (why can't we do what Fixer suggested to fix this?)
13:25 Shara        But using max_lag like this instead is bad
13:25 IhrFussel    paramat, yes °I agree there...we can keep the max_lag but we really need another indictaor for playability then
13:25 Shara        And unfairly favours servers that do not need to run the usual mods due to non-standard gameplay
13:25 IhrFussel    indicator*
13:27 IhrFussel    nerzhul's test server mostly has a max_lag of < 0.1 BUT only cause there is maybe 1 or 2 players + minetest_game only
13:31 paramat      anyway please comment in that PR, your concerns don't need a separate issue
13:32 nerzhul      IhrFussel, yeah stresstest is needed :p
13:32 nerzhul      but after fixing Fixer problem and your black screen problem
13:33 Fixer        IhrFussel: https://github.com/minetest/minetest/issues/6508
13:34 IhrFussel    paramat, my suggestion to add a second lag value to /status will be its own issue
13:34 nerzhul      does someone cancel builds on travis ?
13:38 rubenwardy   server max_lag is a good indicator on how under load a server is
13:38 rubenwardy   however
13:38 rubenwardy   different types of servers require different step sizes
13:39 rubenwardy   CTF would be awful with > 0.05
13:39 rubenwardy   but creatives would probably be ok up to 0.3 or so
13:39 Shara        rubenwardy: problem is max_lag peaks, then stays up
13:39 Fixer        rubenwardy: problem is you don't know CURRENT LAG
13:39 Fixer        we want to know LAG NOW
13:39 Shara        You realise you've been playing on RC and it's never that low?
13:39 Fixer        Not some weird 0.997 avg
13:40 Shara        Yet if RC was less than okay, you or many others would be complaining.
13:41 rubenwardy   what was RC usually?
13:41 Shara        I can be 2+
13:41 Shara        It*
13:41 Shara        Especially when busy
13:41 Shara        The thing is, it's not constantly 2+, but that's what max lag will make it look like
13:42 Shara        If servers are to be ranked on lag, it should be a much more up to date value
13:42 IhrFussel    WAIT WAIT WAIT.... it says "lag": 0.20199376344680786, in the server list for my server, so it CANNOT be the max_lag value
13:43 rubenwardy   servers should be penalised if they regularly go above 1s (eg: minutely), but not if they went above a second 15 minutes ago
13:43 Shara        (Fussel, thsi is why I asked clarification on meaning earlier, but no one has said my assumption is wrong)
13:43 rubenwardy   IhrFussel, *1000
13:43 rubenwardy   !math calc 0.201*1000
13:43 ShadowBot    rubenwardy: 201
13:44 IhrFussel    201 ms but max_lag right now above 1.5
13:44 rubenwardy   huh
13:44 Krock        https://github.com/minetest/minetest/blob/master/src/server.cpp#L559
13:44 Krock        because lag != max_lag
13:45 IhrFussel    You simply take the current server step time and send it to the server
13:45 rubenwardy   yeah, I meant m_lag
13:45 Krock        the lag reported to the server list is an actual average lag value of 100 measures
13:45 IhrFussel    list
13:45 rubenwardy   100 measures, so a bit less than 100 ticks to stablise?
13:45 IhrFussel    Then it's so high cause I set my server_step to be 0.2 instead of 0.1
13:46 sfan5        rubenwardy: it would be easier to fix the max_lag calculation not to include lagspikes so heavily
13:46 IhrFussel    So you penalize servers that don't need such a high update frequency
13:47 IhrFussel    0.1 is too much for my machine but 0.2 is still very playable... so do you want to FORCE server owners to set server_step to 0.1?
13:47 rubenwardy   all servers which use chests and doors need < 0.5 step, as that's how long you'd need to wait for something to happen due to no client side prediction
13:48 rubenwardy   and 0.2 is fine for a creative server
13:49 IhrFussel    It should also be fine for a non-pvp-focused survival server
13:50 rubenwardy   if they don't have mobs then yes, they're basically equivilent to a creative server but with hunger/health
13:50 IhrFussel    Where's the difference in crafting/mining/building/chatting between 0.1 vs 0.2?
13:51 IhrFussel    I'd say even with mobs (like mine) it's okay... you don't need realtime action to fight those simple mobs this is not WOW
13:51 rubenwardy   I disagree, entity movement is annoying > 0.12 or so
13:52 IhrFussel    We are talking about "playability" and 0.2 steps make a server with mobs still playable
13:53 IhrFussel    Either that or myself and all my active players have an extremely low requirement for playability
13:55 IhrFussel    Well but since we actually know now what the "lag" value in the server list really means (an actual average over time) I'm less against merging the PR
13:56 rubenwardy   https://github.com/minetest/minetest/pull/6504#issuecomment-334936784
13:57 IhrFussel    The only problem is that some servers set a different server_step value and if the PR can take that into account
13:58 rubenwardy   that shouldn't matter really. The quality of multiplayer is linked to the step size and the variation of step size
13:58 rubenwardy   although actually
13:58 rubenwardy   as said different subgames require different step sizes
13:59 paramat      Shara and server owners please can you express your need for an option in https://github.com/minetest/minetest/pull/6490 so that it doesn't look like just me requesting it?
13:59 IhrFussel    I'd be fine with those values...my server would be yellow then and maybe I'd decide to switch to 0.15 steps to make it green
14:00 IhrFussel    green-yellow*
14:00 rubenwardy   0.201 would be green-yellow too
14:03 rubenwardy   the point of this should be to highlight servers which don't have lag spikes and have a consistent low server step
14:03 rubenwardy   so 0.2 is way too low as a value
14:04 rubenwardy   I might get rid of green-yellow in my table, and combine it with green
14:04 Shara        I think things you have in green-yellow should be green
14:05 rubenwardy   done
14:05 IhrFussel    rubenwardy, the values the server list reports are actually pretty close to the ones my own avg_lag reports too (between 0.20 - 0.30 depending on load)
14:05 IhrFussel    mod*
14:06 Shara        I'm wondering if it's just that sofar has said max_lag instead of m_lag but... I don't know
14:12 Fixer        PROMISE ME THE LIGHT
14:15 * rubenwardy checks out sauth
14:15 nerzhul      Fixer, are you around ?
14:15 IhrFussel    Not sure but m_ in c++ usually just indicates that it's a member var or something like that
14:15 Fixer        nerzhul: yes
14:16 nerzhul      i need you to help me
14:16 rubenwardy   shivajiva: why haven't you released sban/sauth to the server? They're very useful
14:16 nerzhul      to understand your bug
14:16 rubenwardy   should I be using them on a production server? :D
14:16 Shara        They can be used
14:17 Shara        I'll probably switch to sauth soon given what happened today
14:17 rubenwardy   also, "It's not required for multiplayer games" -> don't you mean singleplayer?
14:18 nerzhul      IhrFussel, Fixer : i push force my branch, i reduced the commit number
14:19 nerzhul      i rebase it too
14:22 Fixer        nerzhul: so how can I help you?
14:22 nerzhul      i'm thinking :p
14:23 nerzhul      the first thing i want to understand is why you are disconnected
14:23 nerzhul      can you capture some packets on your machine ?
14:24 shivajiva    ruben you are correct it's not clear I'm referring to the auth,txt, released to server? You mean minetest mods I guess, wanted a bit more ground testing and feedback
14:25 nerzhul      i'm interested to have a network capture (with wireshark for example as you are on windows)
14:25 nerzhul      Fixer, you have the same problem on my server right ?
14:25 Fixer        nerzhul: kinda same
14:25 rubenwardy   *to the forums
14:25 nerzhul      Fixer, nice it's easier then
14:26 shivajiva    ah, yea I should
14:26 nerzhul      Fixer, can you connect on my server one time now i need to gather your ip and kill the client ?
14:26 Fixer        kill the client?
14:26 shivajiva    seems like they should be fused together though?
14:26 Fixer        it crashes anyway
14:27 nerzhul      yeah maybe :p
14:27 nerzhul      i just need you to connect once first to gather your IP, can you ?
14:27 nerzhul      (with the latest code)
14:27 Fixer        nerzhul: done
14:28 Fixer        nerzhul: tried to connect
14:28 nerzhul      i see
14:28 nerzhul      let me a second
14:28 nerzhul      ok try it another time with freshly started client
14:29 nerzhul      and tell me when it's done
14:29 Fixer        nerzhul: done
14:29 nerzhul      ty
14:29 nerzhul      i have the capture i can analyse
14:34 nerzhul      Fixer, wow
14:35 nerzhul      bad checksum on your TCP packets
14:39 nerzhul      sorry on server TCP packet
14:39 nerzhul      very strange
14:41 Fixer        kekek
14:48 IhrFussel    paramat, commented but looks like we are in the minority (I'm neutral too by the way but all for fairness)
14:49 IhrFussel    I mean neutral as in "it doesn't affect me"
14:50 * Shara      changes server step on RC to 0.1 again just to see what happens
14:51 nerzhul      Fixer, can you do another test i missed some packet data in the capture
14:51 paramat      Shara Megaf benrob0329 rubenwardy i tried arguing for an option in https://github.com/minetest/minetest/pull/6490 'Correct smooth lighting at covered corners' but have had enough now, it's up to others now to argue for it, i'm neutral. it looks very likely an option will not be added
14:51 Fixer        yes
14:51 nerzhul      Shara, you will just increase CPU usage
14:51 Fixer        nerzhul: ready?
14:51 nerzhul      yes
14:51 nerzhul      look at appletree for a good step :p
14:51 nerzhul          0.025
14:52 Fixer        done
14:52 Shara        paramat: so more is broken for players again
14:52 rubenwardy   CTF is consistently at 0.05, the dedicated_server_step. Might try halving it again
14:52 Shara        nerzhul: I will also report lower in lag, which is now possibly quite important
14:52 nerzhul      it's not real
14:53 Shara        I know my server does not need this, but if the public list PR is accepted...
14:53 nerzhul      lag depend on server step loop, if you have many long action you will just increase it if they doesn't depend on timers
14:53 Shara        Well, I have cpu to spare
14:53 Shara        RC used to report 100 or 99 lag way back... I changed server step and it reported 148
14:53 Shara        Yet gameplay was actually better
14:54 Shara        Now if you're adding indicators of how good a server is based on this value...
14:54 Shara        I need value to be lower
14:54 IhrFussel    Shara, cause 0.2 steps means automatically "0.2 sec lag per step"
14:54 Shara        RC now reports 100 lag instead of 148
14:55 nerzhul      Fixer, can you do another test ? i'm ready
14:55 Fixer        yes
14:55 IhrFussel    Yes the value the server list reports is HIGHLY dependend on the server_step setting which is why that's not fair still
14:55 Fixer        done
14:55 Shara        Totally not fair. :)
14:57 nerzhul      Fixer, it seems each time server disconnects you when pushing meshes
14:59 rubenwardy   lol
14:59 rubenwardy   remote media is massively triggering rate limiting on my server
15:00 nerzhul      Fixer, another test ? i added a crash on the disconnection to have the backtrace :p
15:00 Fixer        done
15:01 nerzhul      okay you are disconnected due to packet sending error
15:02 IhrFussel    Shara, if the server_step is set to 0.2 and the previous step took <= 0.2 secs then it would usually mean ZERO lag
15:04 Shara        When my server was on the default, it reported 100 lag. I switched to .015 (it improved some problems I had back then) and it reported 148 lag. These values were pretty consistent in most cases
15:04 Shara        Now it's back to default and again reports 100
15:05 IhrFussel    Yes but if the step didn't take longer than the set value in minetest.conf then I don't consider it to be "lag"
15:05 Shara        But I think I won't leave it here, as for some reason CPU use goes up disproportionately for me
15:05 IhrFussel    Cause it "ticked" at the correct speed
15:05 Shara        Well, neither do I, but I am not the one in charge of the server list :)
15:06 IhrFussel    The lag calculation should be PER SERVER and take the server_step setting into account
15:06 IhrFussel    And not just assume all servers run at 0.1 steps
15:07 Shara        Well, there should also probably be somethign to check server_step isn't some ridiculous value
15:07 Shara        Since that alone could make a server quite unplayable
15:08 IhrFussel    Yes that too... but 0.2 steps is playable, no question
15:08 nerzhul      Fixer, can you do another test and update your git tree, i added more logs on the disconnection
15:09 Fixer        nerzhul: you want test now, or after git tree update?
15:10 IhrFussel    I had my server on 0.4 for some time in the past and the only thing I noticed lagging horribly were the player model updates since I use playeranim
15:10 nerzhul      yeah compile the new version and connect
15:10 nerzhul      i should have interesting logs telling me why there are a sending error
15:12 Fixer        okay
15:14 IhrFussel    Sometimes I just think core devs are focused too much on singleplayer and they have little understanding about how long-term servers perform/what they need
15:15 nerzhul      IhrFussel, it depend on who, every coredev has his own roadmap
15:15 nerzhul      for me the light bug should be fixed, it's just definitively a bug, i understand players are using workaround, but they can fix it
15:15 nerzhul      IhrFussel, you seems to have an example usecase, show us :)
15:16 IhrFussel    nerzhul, it's not just that...for example (most) core devs thinking that a max_lag on servers of 2-3 is "critical" just shows me they rarely play on servers it seems
15:17 IhrFussel    I'm talking about things that you simply can't experience in singleplaayer...you have to connect regularly to servers to understand them
15:17 nerzhul      it's critical to have the more accurate value to reflect to users, but it doesn't block mt work
15:19 IhrFussel    Yes but there were many more things the majority of server owners disliked but still was merged
15:19 IhrFussel    Because (most) core devs cannot think like server owners
15:20 IhrFussel    They have other priorities
15:21 nerzhul      it's not always easy to make good decisions, the lag decision should not be based on the idea that bad laggy server are currently green in serverlist whereas they are unplayable
15:21 nerzhul      Fixer, do you recompile the whole project each time ? :p
15:21 paramat      '(most) core devs thinking that a max_lag on servers of 2-3 is "critical"' not 'most', and it depends how often that kind of extreme lag happens
15:21 Fixer        yes, i will modify the script later
15:22 paramat      i think your players do have a low standard because you admit you have many long lags
15:22 IhrFussel    nerzhul, do you consider a 0.2 timeframe to be "lag"? I'm pretty sure 99% of players can't even tell the difference in gameplay in most situations
15:22 nerzhul      200 ms is laggy, yellow lag
15:23 nerzhul      to show you a concret example, in World of warcraft lag is yellow after 100ms and red after 500ms
15:23 IhrFussel    But if the server_step is set to 0.2 it would mean internally there is *no* lag
15:23 paramat      core devs actually have server's and player's needs as a high priority, higher than you think, but sometimes get it wrong of course
15:24 IhrFussel    Why would the intended tick rate be "lag" is what I wonder
15:24 Fixer        in my experience, for online fps play you need at least 50 ms, 100ms - starts to suck, 150ms - sucks/unplayable
15:25 paramat      but WOW probably doesn't have as much variability as MT with lua mods running, so we have to be more lenient
15:26 paramat      server step should be included i think, as that adds to the experienced lag, otherwise a server could set server step to 1s and still get a green rating
15:26 Shara        Fixer: yea, but that's FPS servers.
15:26 nerzhul      Fixer, erf i got a crash in my logs, i will play with my wife some minutes, will fix it later
15:26 Shara        Why should a creative be marked less than green for not meeting FPS standards?
15:27 Fixer        minetest is first person sh...something
15:27 paramat      yeah
15:27 paramat      (yeah to Shara's comment)
15:27 Shara        A CTF server needs to meet that standard, yes.
15:27 Fixer        nerzhul: compile done
15:27 Shara        My creative doesn't.
15:28 Shara        Light survival servers without a massive focus on mobs wouldn't need it either.
15:29 IhrFussel    My survival server that's not focused on PvP works too and nobody complains...so why should I need to reduce the step time to 0.1 if 0.2 works? And I still get a penalty for it?
15:29 Shara        I never had issues on your server Fussel (unless your moderators were playing "let's WE crazy large areas at once")
15:30 Shara        And that was only really bad the one time. :)
15:30 IhrFussel    Just make the server send their server_step time to the list and take it into calculation... you could make it so >= 0.3 steps is ignored for example
15:31 Shara        I'm going to leave RC on 0.1 fo r awhile anyway, just to see how it holds up
15:31 rubenwardy   woo! remote media now works with Nginx
15:31 IhrFussel    You honestly created this issue yourself core devs... I mean you added the option to adjust the server_step and now you say "if you set it higher you will have a disadvantage"
15:32 rubenwardy   https://gist.github.com/rubenwardy/ab93c2ec5947d36af2e8c4c5e4e5f03f
15:33 sfan5        why would you do rate limiting?
15:33 rubenwardy   to stop abuse
15:33 Shara        Can't avoid things getting implemented seperately and then adding up to make a problem somewhere. So no blame on the devs there. It's just about whether it gets taken into consideration or that PR is merged as is.
15:33 rubenwardy   although they're lots of small files, tbf
15:33 IhrFussel    Why even implement that option then in the first place? If the option wouldn't have existed I would've WATCHED the server performance more as I added new mods...but since I though nobody will complain about 0.2 that's what I based everything on
15:34 rubenwardy   :/
15:34 rubenwardy   because different servers can get away with different step sizes, so it doesn't make sense to have no option
15:35 IhrFussel    But it makes sense to penalize those different servers?
15:35 paramat      IhrFussel your serverstep counts because it adds to experienced lag, as i just wrote above
15:36 IhrFussel    Let's be real guys there is no huge difference between 0.1 and 0.2 steps in server gameplay
15:36 Shara        paramat: just how to balance it? On one hand, reducing server step on a server where being able to shot at players matters would in itself be an issue, but on a creative... There's no negative impact here
15:37 IhrFussel    Some players might even live so far away that their network lag exceeds the server step time that's why I find a server_step of 0.1 not that important
15:38 Shara        Should the server list be strict, in that it measures based on what the highest requirements for a good server might be (CTF for example), or more relaxed, where it measures based on a server with lower requirements? (probably creative)
15:38 Shara        ?
15:38 IhrFussel    In fact it means basically I make it a lot more fair to players with slow ping
15:42 paramat      Shara true, but these are small changes in server step: 0.1 vs 0.2, i don't agree with marking those using 0.2 as 'less playable', i'm arguing for more leniency
15:42 IhrFussel    Shara, such a strict rule would basically "disallow" mixed servers... like my server is PvP enabled (according to the server list) but in reality it's completely optional-per-player and doesn't rely on realtime actions
15:43 paramat      .. so more relaxed
15:43 Shara        IhrFussel: it would not disallow anything, but it would lower the rating of any server, regardless of type, not meeting that strict standard
15:43 Shara        But I'd support what paramat just said
15:43 IhrFussel    Yes my server is survival but relaxing survival...I'd call it "social survival"
15:44 Shara        paramat: thing is, it's not just about server step, but the impact it has on reported lag
15:44 Shara        IhrFussel: Well, you know I was running round there killing things with no armour, so yes, I'd say very relaxed :)
15:44 paramat      anyway i have asked we reconsider what value should be used and at what values each colour is shown
15:47 paramat      lag spies tend to be high, 3+s, so the server step being 0.1 or 0.2 makes little difference to that
15:47 IhrFussel    I said I'd be fine with "yellow-green" when step is set to 0.2 but such a measurement would depend a lot on that setting
15:47 paramat      hehe *spikes
15:47 nerzhul      Fixer, i pushed the logging fix
15:47 paramat      a server with 0.2 step should be able to get green certainly
15:48 paramat      which is why i disagree with sofar's initial PR
15:50 Shara        Hmm
15:50 Shara        From the PR: "All english spoken servers are likely focussed more on US player base"
15:50 Shara        Total nonsense
15:50 IhrFussel    I can only tell that my avg lag mod mostly reports 0.2-0.21 if there is no max_lag higher than 0.5 secs ... when more players connect it can go up to 0.22 - 0.25 which tells me there have been a few little spikes and sometimes it reaches 0.3 which is critical so such an indication makes sense
15:51 Shara        Most servers I spent any time on had english speaking players for all parts of the world
15:51 Shara        Though I do still agree ping isn't a great measure either way
15:52 IhrFussel    About the ping...not sure if it would be good to ping the server directly (c++ code abuse possible)
15:53 rubenwardy   it's bad for privacy
15:53 rubenwardy   if we need ping, it would be better to host serverlist-style servers around the world, and have them ping each server. Players then connect to the closest one based on IP
15:54 IhrFussel    People then could also re-code their client to bomb a server with ping requests
15:54 rubenwardy   or maybe just have them report the ping to the serverlist, and the serverlist serve    ping = { uk = 0.2, us = 0.4 }   etc
15:54 rubenwardy   they could do that already
15:55 rubenwardy   there's even a very easy python script for it
15:55 IhrFussel    I mean they already can but if the "ping directly to server" code is implemented they likely just need to adjust the interval anymore
15:56 Krock        rubenwardy, people from asia or australia won't be happy
15:56 Krock        "that number is unreliable for us"
15:57 rubenwardy   then have asian and australian servers, and show nothing if there is no near ping server. This is a lot of infrastructure though, so I'd say it's not worth it
15:57 Krock        and minetest isn't known enough to have such a big network
15:57 IhrFussel    How do other big games do it? They probably have not only one server list server worldwide
15:58 Krock        ping each server client-sided
16:00 IhrFussel    Oh yeah...since most games aren't open source the code is safe and can't be abused
16:00 rubenwardy   that's not the problem
16:00 rubenwardy   the problem is privacy
16:01 rubenwardy   it's already trivial for someone to ping a server
16:01 IhrFussel    Why privacy? What else does it tell other than the time it took?
16:01 rubenwardy   especially for the sort of people who want to abuse networking
16:01 rubenwardy   every single person that opens up Minetest would have their IPs given to all servers
16:02 IhrFussel    And the big game companies don't care?
16:02 rubenwardy   yeah
16:02 shivajiva    first one - https://forum.minetest.net/viewtopic.php?f=9&t=18603
16:02 rubenwardy   wooow
16:03 rubenwardy   oop
16:03 rubenwardy   +s
16:03 IhrFussel    And connected to a server the client already pings the server directly right?
16:03 rubenwardy   *woooo
16:03 rubenwardy   I should probably wait until I need to ban someone before making a comment
16:03 shivajiva    impressive post count huh
16:04 Fixer        nerzhul: failed to connect to your server
16:04 shivajiva    yea, good idea ruben :)
16:04 rubenwardy   also, pinging every server would be sllooow
16:04 nerzhul      Fixer, i forgot to start it :) it's done
16:04 Fixer        lol
16:04 Shara        I've ban lot sof people with sban so far. Works fine :)
16:05 Fixer        nerzhul: "connection lost"
16:05 Shara        Regular and temp bans both tested
16:06 shivajiva    better the client gives the ping and a distance would be awesome for actually reporting something meaningful
16:07 IhrFussel    Wouldn't that require a geoip lookup?
16:07 Calinou      clients should just ping servers directly, but it's just me :P
16:07 jas_         IhrFussel: ping me
16:07 Calinou      it works like this in every other game
16:07 jas_         tell me how many ms
16:07 jas_         there's your number
16:07 Calinou      that would require a lot of refactoring, though
16:08 Calinou      the server list formspec would also need to be able to update asynchronously
16:08 IhrFussel    0.174 from Germany
16:08 jas_         well, in certain older fps, you had to hit the refresh button (or hit space hehe)
16:08 shivajiva    Calinou is correct
16:08 jas_         greetings from maine
16:09 IhrFussel    But to get the location and calculate the distance requires geoip I think
16:09 Calinou      jas_: it has to be done asynchronously, so that you don't need to wait for all servers to be pinged to see the list
16:09 Calinou      else, it'll take much more time than needed
16:09 Calinou      Cube/Cube 2 games do it right
16:09 Calinou      in fact, all server information is gathered from the server directly, IIRC
16:09 Calinou      not from the master server (it only has a list of IPs)
16:10 jas_         well if you're going to do it, you may as well do it right
16:11 IhrFussel    jas_, trying to remember where maine is in the world xP
16:12 jas_         it's the head of the cow
16:12 jas_         oh wait.  nevermind that.  it's eastern us
16:12 shivajiva    http://www.ip2location.com/tutorials/find-distance-between-2-ips-using-bash
16:15 IhrFussel    Ah so USA ... quite a large ping
16:15 IhrFussel    But the largest would be from Japan I guess
16:27 jas_         nerzhul: did you fix it?
16:28 jas_         you wanna try me hosting and you connecting?
16:33 nerzhul      jas_, no thanks :p
16:35 jas_         ok i'm putting it on dcbl.duckdns.org:30002 anyway
16:37 Krock        can you become a git master by git checkout master?
16:37 jas_         i think that's right
16:39 Calinou      git checkout username
16:39 Calinou      "username checks out", as Reddit says
16:57 nerzhul      Fixer, can you test another time please ? :)
16:57 nerzhul      i really want to handle this crash properly :p
16:57 nerzhul      this log*
17:00 Hijiri       well, #5612 and #5819 are waiting for one approval if someone wants to eliminate them
17:00 Krock        https://github.com/minetest/minetest/pull/5612 - Allow overriding tool capabilities through itemstack metadata
17:01 Krock        https://github.com/minetest/minetest/pull/5819 - Fix default item callbacks to work with nil users
17:01 Hijiri       bot broken today?
17:01 * Krock      is a slow bot
17:01 Krock        Hijiri, merely wrong channel
17:01 Hijiri       oh
17:01 nerzhul      bot doesn't answer on -hub it's not the dev chan
17:02 nerzhul      jas_, can you connect to my server ? if you have same problem as Fixer it's nice for me
17:02 Hijiri       yeah, I meant to post in #minetest-dev since paramat was talking about rising issue counts
17:02 Hijiri       but if it's here it will probably be seen anywa
17:02 nerzhul      just have an up-to-date tree
17:19 shivajiva    second one - https://forum.minetest.net/viewtopic.php?f=9&t=18604
17:22 Fixer        nerzhul: i'm back
17:23 nerzhul      Fixer, nice, test when you want
17:23 Fixer        nerzhul: will test now
17:23 ThomasMonroe hey Fixer what was that fix for the item frame problem in mesecons?
17:23 Fixer        nerzhul: done
17:24 rubenwardy   bot should probably answer in this channel, as it's supposed to be a dev-server-community channel or w/e
17:24 Fixer        ThomasMonroe: no idea..
17:24 ThomasMonroe how do you make it not pushable?
17:24 ThomasMonroe wait, maybe i can copy some code somewhere....
17:28 nerzhul      Fixer, connection reset by peer is the error
17:28 nerzhul      Fixer, can that be due to a firewall/antivirus ?
17:29 jas_         2017-10-07 13:29:09: ERROR[ClientConnectionThread]: Failed to read packet headerfrom 62.4.22.5:30001, disconnecting user. Error: End of file
17:30 jas_         connection lost
17:30 nerzhul      yeah we now have logs in both sides
17:30 nerzhul      2017-10-07 19:29:08: ERROR[ServerConnectionThread]: Failed to send packet to 172.101.13.131:33970, disconnecting user. Error: Can't assign requested address
17:30 nerzhul      here is yours jas_
17:30 nerzhul      Can't assign requested address, very strange
17:30 jas_         weird i don't even see your attempt
17:31 jas_         welp, gn o/
17:31 ThomasMonroe Fixer, i found it, i just add: mesecon.register_mvps_stopper("itemframes:frames")
17:36 sofar        I see the flamewar has started
17:36 sofar        we can just remove the ping and not put anything back
17:36 sofar        and it would be an improvement
17:39 Shara        That I'd agree with
17:40 nerzhul      Fixer, i added more accurate logs, can you test ? also when the test finish i think you can have some interesting client error logs
17:40 Shara        Though it would be nice if there was something genuinely better
17:40 Fixer        nerzhul: connect to your server?
17:40 sofar        max_lag is actually a really good indicator
17:40 Fixer        nerzhul: do i need recompile?
17:40 sofar        why shouldn't lag spikes like WE matter?
17:40 sofar        they do affect gameplay
17:41 Shara        That's not why I disagree with it.
17:41 sofar        still reading the thread, lol
17:41 nerzhul      Fixer, it can be better as the logs are also cleint side
17:42 Shara        Hmm, I guess 0.1 server step is fine on RC anyway. CPU isn't maxing out with 20 players.
17:42 Fixer        recompiling
17:43 nerzhul      Shara, just use default, 0.09
17:43 sofar        many servers are at 0.05 btw
17:44 sofar        I'm going to collect a shit ton of data and make some actual statistical meaningful statements
17:44 Shara        Performance for players was already totally acceptabe at .15
17:44 sofar        might take me 2 days
17:44 sofar        btw I have no issues with adjusting the scale or even the colors
17:45 Shara        sofar: in fairness, I'll probably be fine with it if you relax it
17:45 sofar        yeah
17:45 Shara        I would really prefer a better measure in general though
17:45 sofar        but like i said, there's only about 15 or so servers that are not "green" with this range
17:45 Shara        The thing is, some kinds of server don't need to be as fast as others
17:46 sofar        which is fine, players can decide for themselves
17:46 Shara        sofar: yea, but guessing one of mine is included, and that's a bit annoying, everything considered
17:46 sofar        that's the thing, now that the icon actually means something, players can make up their minds better
17:46 sofar        with the ping icon, that was impossible
17:47 Shara        sofar: maybe you can help me understand this better.
17:47 Shara        If a server reports around 5.8 max_lag every time I actually check it, yet is showing arround 220 on the server list.. which value are you looking at?
17:49 sofar        It's just using the value that servers send and have been sending, so I gotta trace down in code first what that value actually is (doing so, now)
17:51 Shara        RC is reporting max_lag=0.621458, with 21 players, and 100 for lag on the list, which feels fine
17:51 Shara        But then DL reports max_lag=5.62744 and shows 224 on the list
17:51 Shara        So I don't really understand the relationship between the two lag values
17:51 sofar        src/server.cpp: m_lag
17:52 * Shara      reads
17:53 nerzhul      Fixer, tell me when you tested
18:02 Fixer        i'm back again
18:03 Fixer        nerzhul: done
18:04 nerzhul      Fixer, do you have a firewall enabled or an antivirus ?
18:04 nerzhul      for me you closed connection
18:04 nerzhul      is there any MT interesting error logs on your side ?
18:06 nerzhul      Fixer, ^
18:07 Fixer        nerzhul: those https://pastebin.com/raw/7B92tSp9 (warning i did some translations, could be wrong)
18:07 Fixer        nerzhul: i have firewall but minetest is enabled in it
18:07 nerzhul      nice Fixer it's the problem
18:07 nerzhul      the UDP sending failure make your client disconnect
18:08 nerzhul      Fixer, i know what is the problem, i provide a patch in 5 minutes
18:09 Fixer        ok
18:13 nerzhul      Fixer, i pushed the fix, it's pure client side, you ar e free to test it asazp
18:14 Fixer        ok
18:20 sofar        https://i.imgur.com/45iGmTe.png <- different m_lag visual
18:20 sofar        looks better, not as harsh if you ask me
18:20 Shara        sofar: Seriously, why is RC not green?
18:20 sofar        lemme check data for it
18:21 Shara        It's reporting 99 on the server list right now
18:21 Shara        max_lag=0.684356
18:21 Shara        There is no way a creative server needs more than this
18:21 sofar        0.68 o_O
18:21 sofar        that's... terrible?
18:21 Shara        No, not at all
18:22 rubenwardy   m_lag, remember
18:22 rubenwardy   0.6s would be terrible
18:22 Shara        That's a very low max_lag for  aserver with 20+ players
18:22 Shara        Yup
18:22 sofar        let me show you what the values are on all the servers
18:22 rubenwardy   red cat is currently at 0.99
18:22 rubenwardy   m_lag
18:22 Shara        I know Dark lands has issues currently, but I have no clue what in the world to do to fix them
18:23 rubenwardy   no
18:23 rubenwardy       0.099
18:23 Shara        But if RC is not green, your valkues are wrong
18:23 Shara        values*
18:23 sofar        https://gist.github.com/sofar/c0a2f83368fccc52aee7b2add031b79f
18:23 sofar        these are all current lag values, sorted
18:23 sofar        notice that only the bottom 10 will be labelled red
18:24 sofar        9 will be orange, then
18:24 Shara        That doesn't matter
18:24 Shara        RC should be green.
18:24 sofar        look
18:24 sofar        stop that
18:24 sofar        "my server should be green" is entirely irrelevant
18:25 sofar        please, at least say "m_lag value xxx should be yyy"
18:25 Shara        Any server that is perfectly playable should be green
18:25 sofar        tantrum
18:25 Shara        I use RC as an example because it happens to be the server I monitor most closely.
18:25 sofar        we don't have that data
18:25 sofar        we have a perfectly usable indicator for server responsiveness
18:25 rubenwardy   except her server is 100, so should be in the green under your own algorithm. Probably an old screenshot
18:25 Shara        Not a tantrum at all.
18:25 sofar        I can fetch rc's data, it will take some grepping
18:26 Shara        RC has, righ tnow, 21 players, is #1 in the public list, and displays 99 for lag on the public list
18:26 sofar        I don't know what "99 lag" means
18:26 sofar        99 seconds?
18:26 Shara        That tells me it performs quite well under all circumstances
18:27 Shara        Very few servers show lower than 99
18:27 rubenwardy   sofar, / 1000
18:27 rubenwardy   so 0.099
18:27 sofar        lets find the value in the data, like real scientists?
18:27 Shara        Your reds seem to be thos ethat show as 200+ on the server list
18:27 sofar        0.205+, yes
18:27 Shara        Red to me would sifnify unplayable, which also might be a bit harsh at thos evalues
18:27 Shara        signify*
18:28 sofar        I was already stating that we can adjust the scale
18:29 rubenwardy   Shara, red cat is green for me with that PR
18:30 rubenwardy   the screenshot is out of date
18:30 Shara        Anyway, you'll note I reported all the values and didn't just winge about my server being too low with no factual data, so I'd appreciate you drop the accusations of tantrums.
18:30 sofar        "lag": 0.09997858107089996
18:30 Shara        rubenwardy: thanks. Good to know.
18:30 rubenwardy   ha!
18:30 sofar        it's green now
18:30 rubenwardy   the server list appears to search case sensitively
18:30 sofar        it was 100% red when I took the screenshot
18:31 Shara        But unless there is somethign wrong with the values currently displaying on the server list, I think anything not too far above 100 should be green
18:31 Shara        What's 100% red?
18:31 rubenwardy   nevermind
18:31 sofar        over .205s
18:31 sofar        green is under .105s
18:31 sofar        about 85% is green atm
18:32 Shara        RC above that would be... odd. But not too important.
18:32 sofar        I've been watching my own server, and it hobbles up and down much more than i previously expected
18:32 Shara        99 would be norm for the default server step, from what I understand
18:33 sofar        there's only 10 servers with a server step < 0.1
18:33 sofar        wait
18:33 sofar        there's only 10 servers with a server step < 0.09
18:33 Shara        But there can be fluxuation.
18:33 Shara        RC had .15 until today, but I might change it back.
18:33 sofar        there's 5 servers with a 0.9 server step
18:33 IhrFussel    sofar, I refuse to have my server labeled as "unplayable" when the avg lag is a bit over 0.20 cmon now
18:33 Shara        It's a creative, so it didn't need more
18:34 sofar        IhrFussel: then you really hate the current ping graphs
18:34 IhrFussel    Do you even know what unplayable means?
18:34 sofar        because they label any non-EU server as unplayable
18:34 Shara        sofar: we're not arguing in favour of keeping ping as a rating.
18:34 IhrFussel    That's another issue the ping doesn't mean as much to players as lag I'd say
18:34 sofar        besides, it doesn't mean that "red" equals "unplayable"
18:34 sofar        that's your interpretation
18:35 Shara        People will read it that way.
18:35 IhrFussel    Red for < 0.3s is wrong no matter how you see it... 0.3s means there were a few larger spikes but not constantly high lag
18:35 sofar        do I have to repeat myself again?
18:35 sofar        <sofar> I was already stating that we can adjust the scale
18:35 Fixer        nerzhul: "connection lost"
18:35 sofar        I just want people to look at the actual data first
18:35 Shara        You sometimes see servers on the public list showing 500+Those would be red I think.
18:36 sofar        https://gist.github.com/sofar/c0a2f83368fccc52aee7b2add031b79f
18:36 sofar        where should the cutoff be?
18:36 sofar        0.5sec? -> only 1 server goes over, so entirely meaningless to even use red
18:36 Shara        No, not meaningless
18:36 sofar        well, 2
18:36 Fixer        i can see that majority is under 0.11
18:37 sofar        0.25sec for red? 5 servers
18:37 Shara        But red should be reserved for somethign really poor
18:37 sofar        sure, fair enough
18:37 Shara        To me, red is like a warning to keep out
18:37 sofar        mind you, there's 200 servers, so 1% red?
18:37 IhrFussel    I'd make everything < 0.3 at least yellow
18:37 Shara        How many of them in active use?
18:37 sofar        is 1% acceptable to label as "red"?
18:38 sofar        that's 2 servers
18:38 Shara        I don't thinkw e should think about it like that
18:38 sofar        sure, active, that's a good point
18:38 Hijiri       "lag" is how long it takes to process a server step right?
18:38 Shara        We should think about it in terms of what's actually playable
18:38 Hijiri       or some statistic of it
18:38 IhrFussel    Hijiri, yes and the avg is all step times combines devided by the amount (the server list divides it by 100 I think)
18:39 Shara        If you have a server flagged as red which is highly active and players don't have an issue... it makes me think it shouldn't be red, but orange or yellow, but probably need to actually look at player activity on servers getting these values
18:39 IhrFussel    divided*
18:39 sofar        so, data wise, there's 2-3 busy servers with 0.2s+ values
18:40 sofar        I wouln't think that labelling them as yellow is unreasonable, or even orange
18:40 sofar        one has 0.5 with 3 clients
18:40 Shara        yellow to orange, yes, but not red perhaps
18:40 Shara        Hmmm
18:40 Shara        A thought:
18:40 Shara        probably not a great one but...
18:40 Shara        Can we take numbe rof players into consideration somehow?
18:40 sofar        a 0.3s with 3 clients...
18:40 Shara        Yea... 0.3 with 3.. ekk.
18:41 Shara        But 0.3 with 15 or 20+...
18:41 Shara        It's different
18:41 sofar        seems around .25 is a reasonable cutoff for reds, if you ask me
18:41 nerzhul      Fixer, client logs ?
18:41 Fixer        nerzhul: same error
18:41 Shara        I'd be more toward .3 myself I think
18:41 rubenwardy   the thing is red is seen as quite bad
18:41 Shara        Or higher
18:42 rubenwardy   like, game breaking lag
18:42 Shara        Yea...
18:42 rubenwardy   technically playable but with much latency
18:42 sofar        .275 and it's a deal?
18:42 Shara        And DL is quite playable
18:42 Shara        But it's fluxuation upwards into that.
18:42 Shara        fluxuating*
18:42 nerzhul      on which commit are you ?
18:42 Shara        And Fussel's server has been quite playable when I've been there
18:43 IhrFussel    My server's average pendles between 0.2 - 0.3 so 0.275 would be fine
18:43 Shara        I'm just scared red is going to be seen as unplayable
18:44 sofar        can I be honest here?
18:44 sofar        people who run a "red" server should be encouraged to get their players a better server
18:44 sofar        but if they can't... that's their choice
18:46 IhrFussel    Are you telling me I should pay more than $25 monthly ?
18:46 Fixer        nerzhul: 60d
18:46 IhrFussel    I think 25 dollars is already a lot for a MT server
18:46 sofar        sure, I pay a similar amount
18:47 IhrFussel    It's actually more expsnsive than my local connection
18:47 sofar        I won't even have internet in half a year
18:47 sofar        if things don't work out
18:47 rubenwardy   Minetest is quite heavy on hard disks with sqlite
18:48 sofar        yeah, my server won't have that problem
18:48 rubenwardy   does sqlite have a memory cache? I don't think it does
18:48 sofar        everything is mmap()d
18:49 IhrFussel    MT seems to be completely unpredictable when it comes to load/performance... sometimes even ONE player can cause a lag spike cause they build something big with mesecons, or they load an area with lots of mesecons mechanics
18:49 sofar        you can't cache writes .. cuz corruption
18:49 nerzhul      lag spike is generally due to : many mods on server loop / emerge many many block
18:49 rubenwardy   you can mark the data in memory as dirty, and defer the writ
18:49 rubenwardy   *write
18:49 IhrFussel    And other times the max_lag on my server is 1.0 with 10+ players
18:50 nerzhul      mesecons & pipeworks mechanics should be integrated to core, the pipe engine i mean
18:50 rubenwardy   it would be better to run mods on a different thread, and communicate using messages
18:50 nerzhul      i tried to understand it at a point but it was clearly impossible to understand (3 years ago)
18:50 rubenwardy   but that's a considerable amount of work
18:50 nerzhul      rubenwardy, yes, sandbox each mod and use mod channel :p
18:50 nerzhul      and threads is a problem not only for mods but for server loop
18:50 IhrFussel    I already adjusted pretty much all mods that run in globalstep...I added a if timer >= 1 then check to them so that they only run once per second and not every step
18:51 nerzhul      as get_node call map
18:51 nerzhul      for example :p
18:51 rubenwardy   no, I meant between the mod thread and the main thread
18:51 nerzhul      and map is on one thread
18:51 sofar        kill all abms
18:51 rubenwardy   make it completely abstracted from the mods to stop breakage
18:51 nerzhul      asynchronous mods is not possible
18:51 nerzhul      get_node is synchronous
18:51 rubenwardy   why?
18:52 rubenwardy   you can pause the lua environment
18:52 nerzhul      you get the current map, you can lock the map, but for each node it's crazy to have a lock on the map
18:52 rubenwardy   my point would be to stop lua blocking packet send, and entitiy updating
18:52 nerzhul      lua doesn't block packet send
18:52 nerzhul      neither with my network rework nor with current code
18:52 Shara        sofar: believe me, I want to do something about my survival server. But I don't know why the performance dropped. It's not always a case of being unwilling to do anything or not caring.
18:53 nerzhul      NetworkPacket is directly sent to ServerConnectionThread in MT
18:53 nerzhul      and it's queued for making other thread consumed
18:53 rubenwardy   Both writing to disk and big mods stop stuff being sent tot the clietns
18:54 nerzhul      no
18:55 IhrFussel    rubenwardy, that I can confirm... if a mod lags the server for a few secs EVERYTHING stops
18:55 nerzhul      sending new generated mapblocks yes , because it's asynchronous and if you server step is high you will send mapblocks with less speed
18:55 rubenwardy   even if packets are sent asynchronously, you need something to send the packets
18:55 nerzhul      yeah, it's not the same problem :)
18:55 rubenwardy   it's the problem I meant
18:55 nerzhul      it's the MT current problem with being too modular
18:56 nerzhul      some hihg consuming generic mechanism should be re-added to core
18:56 nerzhul      like mesecons/pipeworks core engine
18:56 rubenwardy   MT is hardly modular. We have massive god classes
18:56 nerzhul      or mobs api/pathfinder
18:57 IhrFussel    Why does the chat also depend on mod lag?
18:57 nerzhul      no problem to being modular just inject highly generic interfaces from mods to core sounds reasonable
18:57 nerzhul      IhrFussel, because chat send callbacks to lua
18:57 rubenwardy   because of chat callbacks, IhrFussel
18:57 rubenwardy   lol
18:57 nerzhul      mod channels are not affected by lag :p
18:57 nerzhul      hmmm they are but it's different, they are less affected :p
18:57 sofar        LOL
18:58 nerzhul      i already studied the performance paraellisation on MT during my epixel fork phasis, and it's hardly difficult to make server have a new thread
18:58 IhrFussel    But when I enable my profiler there are only a few mods that show a high avg runtime... I ignore min/max as those tend to be rare
18:58 rubenwardy   not client-to-client, client-to-server still needs a mod to consume and respond. (Just correcting lag to mean server thread congestions)
18:59 nerzhul      rubenwardy, client to client is affected too, but i think if mod channels are hardly used it's easy to make them use a dedicated thread, if the network PR is finished and accepted we can forward a network packet to other than server thread
19:00 Hijiri       is anyone aware of a worldcraft iOS version?
19:00 Hijiri       A user on my server says they are using "Worldcraft(3D)" on an iPhone
19:00 IhrFussel    So if I understand it correctly the engine doesn't actually *lag* it's just the c++ -> lua -> c++ callbacks?
19:01 rubenwardy   yes, Hijiri
19:01 rubenwardy   https://forum.minetest.net/viewtopic.php?f=3&t=16707
19:02 sofar        I am *so* happy that not even freemined can't connect to my server
19:03 sofar        problem solved :D
19:03 rubenwardy   lol, double negative
19:03 IhrFussel    Well I won't be able to reduce lag much since many players want to protect areas which causes a 1-2 sec lag when saving/renaming/deleting
19:03 rubenwardy   your server is very accepting
19:03 rubenwardy   IhrFussel, do you use area store?
19:03 IhrFussel    I use the areas mod but the file stores ~ 7,000 areas
19:03 rubenwardy   hmmm, that probably wouldn't help creating areas though
19:04 rubenwardy   IhrFussel, area store is an engine feature that makes the areas mod faster
19:04 IhrFussel    I think areas uses that...one sec
19:04 rubenwardy   yes
19:04 rubenwardy   but you need to compile minetest with libspatial to get a better implementation
19:04 rubenwardy   however it sounds like an IO problem
19:05 rubenwardy   maybe shivajiva will be able to work on a sareas ;)
19:06 IhrFussel    rubenwardy, the HDD on my dedi peaks at 84 MB/sec ... I don't think that's low
19:07 rubenwardy   as said, an IO problem
19:07 shivajiva    yea, just closing the global namespace access on sban and sauth to sqlite and then I'll take a look at adding that string to my bow *chuckles*
19:08 Shara        Please do :)
19:08 nerzhul      iostat 1
19:08 nerzhul      iotop
19:08 nerzhul      use that
19:09 IhrFussel    MT never uses more than 5 MB /sec in iotop
19:09 rubenwardy   I used those commands to find out that the player database was the problem ^
19:09 rubenwardy   also
19:09 rubenwardy   sudo inotifywait -e modify -e attrib -e move -e create -e delete -m  .
19:09 Hijiri       rubenwardy: that one is android
19:10 rubenwardy   that will show you what files are changing in a folder as it happens
19:10 rubenwardy   Hijiri, no there's iOS too
19:10 sofar        env_meta gets written every 2 seconds or so
19:10 Hijiri       oh
19:10 Hijiri       sorry, I didn't read far down enough
19:10 rubenwardy   sofar, if it's the save interval then is 5.3 s
19:10 sofar        ah
19:10 sofar        well that, then :)
19:12 paramat      i was thinking .25-0.3 for red
19:12 sofar        I put it at .275
19:12 rubenwardy   other helpful tools: jnettop, htop
19:13 Shara        .275 is probably okay.
19:13 paramat      (i mean 0.25-0.3 for the lower limit of red)
19:14 sofar        that's what it is
19:14 IhrFussel    It will show my server as red when the load peaks but that's fine
19:14 sofar        paramat: it's .275 for the lower limit of red
19:15 IhrFussel    But servers with server_step at 0.1 still have it easier...just saying
19:16 IhrFussel    I will have to test if my machine can handle .15 too maybe I'll switch then
19:16 sofar        well I'm glad people are somewhat positive about replacing ping with something else :)
19:16 Shara        I never noticed any issues at .15, and it did ease things up a lot.
19:16 IhrFussel    And to have the step at <0.1 is not recommended by devs
19:17 Shara        So it's probably worth you trying it, Fussel
19:17 sofar        right, it'll lower your overall load
19:17 sofar        well default server_step is now 0.09s
19:17 Shara        I'm testing at .1 for the moment
19:17 IhrFussel    Yeah but 0.05 is too much someone said...MT isn't ready for that yet
19:17 paramat      IhrFussel but you didn't attend to moretrees ABM growing 10-20 large l-system trees at once each with a separate lighting update, as you found out from your mod profiler it's moretrees that is very intensive
19:17 sofar        my server runs at 0.05 but I can affoard it due to the game being aggressively designed to avoid bottlenecks
19:18 IhrFussel    paramat, what should I do? Remove the mod? My players would kill me ... I would adjust its code if it wasn't written so extremely complicated
19:19 IhrFussel    I was thinking about removing everything except for saplings/tree/wood blocks and change those to use nodetimers
19:20 paramat      'only a few mods with high runtime' is irrelevant, the server is as laggy as the laggiest mod (moretrees)
19:23 tenplus1     hi folks
19:23 IhrFussel    hi tenplus1
19:23 Shara        IhrFussel: About moretrees, it gets a lot better if you stop new trees appearing on mapgen
19:23 tenplus1     o/ fussel
19:23 tenplus1     hi shara
19:23 Shara        Hi tenplus1
19:24 Raven262     Hi ten.
19:24 tenplus1     hey raven
19:24 tenplus1     what's new ?
19:24 IhrFussel    I know that paramat but I was just saying out of my ~ 160 mods (which may sound a lot but most of them just add a thing or 2) only ~ 10 seem to cause frequent lag
19:24 Raven262     I made a new box on insidethebox.
19:24 Raven262     Now i found myself playing minetest only on that server. Amazing, truly.
19:24 tenplus1     yeah I read that, Sofar said it'd keep anyone busy for an hour :) is it that hard ?
19:25 Raven262     Yes.
19:25 Raven262     I tried to make it harder than the 51.
19:25 tenplus1     ehehe, gotta try it... what number is it ?
19:25 rubenwardy   is it a bug that Minetest's max_user is one more than the actual maximimum users?
19:25 Raven262       168
19:25 tenplus1     hi ruben
19:25 rubenwardy   minetest doesn't have reserved slots...
19:25 rubenwardy   hey
19:25 Shara        rubenwardy: it's just the closer it gets to max, the more likely it is players can't connect
19:25 * tenplus1   had an idea
19:25 IhrFussel    There is a new callback mods can call to still let players with certain privs/abilities through
19:26 paramat      IhrFussel erm, the ABM that grows the 'on mapgen' saplings, adjust that to have a higher 'chance' value but a smaller interval so that it doesn't grow so many trees at once
19:26 tenplus1     hi paramat
19:27 IhrFussel    paramat, higher chance and smaller interval is better than higher interval and smaller chance?
19:27 paramat      yes
19:27 paramat      becasue it splits the work into a larger number of smaller operations
19:28 shivajiva    Hiya tenplus
19:28 tenplus1     hi shiva o/
19:28 paramat      more tiny lags are better than a few huge ones
19:28 tenplus1     would it be possible to have a player override flag that draws everything on the client in full brightness ??? e.g.  night vision
19:28 paramat      and of course, use the option to place 'on mapgen' saplings during mapgen, that grow later
19:28 Shara        tenplus1: that would be nice :D
19:29 paramat      .. that option is the default now i think for that mod
19:29 tenplus1     was thinking since we have a map item for overhead maps, a telescope item for zoom priv, that it would be nice to have client draw everything in full brightness if say a special night vision potion was used and then turn it off agter a time
19:32 paramat      Ihr, remember the 'chance' value is inverted in meaning, larger value == less chance
19:33 IhrFussel    paramat, I know right now I set it to interval 8 and chance 10 ... not good?
19:35 CalebDavis   hi ten
19:36 CalebDavis   the full brightness would be awesome and since we will have a map and binoculars it would be good as night vision goggles
19:36 tenplus1     https://github.com/minetest/minetest/issues/6509
19:36 tenplus1     hi caleb
19:37 Krock        > full brightness
19:37 Krock        > totally white
19:37 Krock        "helpful"
19:37 tenplus1     hey krock :)
19:37 Krock        hi tenplus1
19:37 tenplus1     it'd be a neat feature to have available for all sorts of potions and abilities
19:38 tenplus1     hell, even a reverse feature for blindness would be handy ;SD
19:38 Fixer        nerzhul: any news?
19:39 tenplus1     hi fixer
19:39 Fixer        hi tenplus1`
19:45 Fixer        looking at core i3-8100 linux benchmarks, very impressive results
19:45 tenplus1     hi thomas
19:45 Fixer        quad core will be mainstream very soon
19:45 tenplus1     ??
19:45 ThomasMonroe hi ten
19:45 paramat      IhrFussel i would set chance higher than 10. work out roughly how many mod tree saplings can be in the active block areas of an average number of players (10?), this value could be around 128 saplings, you would set the chance to that value to have one tree grown per ABM action. then keep halving the value until you see a problematic amount of lag. interval could be less but would keep it >=2s. maybe start with
19:45 paramat      chance 64?
19:46 Fixer        and that means, minetest should utilise more cores since we are pretty much going into quad core in budget PC
19:46 Fixer        both with AMD and Intel
19:46 tenplus1     sweeet
19:47 Fixer        core i3 is now fully quad core CPU (pretty = core i5)
19:47 paramat      .. and monitor the profiler data for the moretrees ABM to find the optimum 'chance'
19:47 Fixer        core i3 8100 vs my core i3 2120 is 2x speed boost in many applications
19:47 Fixer        for the same price
19:48 Fixer        and i had without incl graphics
19:49 Shara        Looks like max_lag on DL is dropping constantly then goes back up whenever someone joins.
19:49 tenplus1     coudl just be loading the area around them when they join shara
19:49 Shara        So when people are joining all the time...
19:50 Shara        tenplus1: it's too much for that
19:50 Shara        DL was always a bit slower than I'd have liked, but it's gotten steadily worse despite me not changing anything.
19:52 Shara        If I get a break in people joining, the performance is okay
19:52 Fixer        and who is joining? those usual HitlerXXX who then stand on spawn or time out after joining
19:53 Shara        Fixer: None of those currently.
19:53 Shara        Not a single number named player yet
19:54 Shara        At 8 players and max_lay = 3.6 currently, but it spikes to 6, or close, anytime someone joins
19:54 Shara        Dropping steadily all the time no one new is joining
19:54 Shara        Yup.. a join and back to 5.4
19:55 tenplus1     must be loading something in the background per player join that's causing issue
19:55 Shara        That's all I can think of
19:56 Shara        The question is what
19:56 Fixer        phoronix in 2017 - still no good conclusions page
20:04 IhrFussel    paramat, thanks for those suggestions, I changed the values now based on your input ... and I set the server_step to 0.15 let's see if the machine can handle it fine
20:09 IhrFussel    sofar, would 0.15 get a green indicator?
20:11 Shara        IhrFussel: if it doesn't have more problems that push the score, then I think so
20:13 IhrFussel    Well right now it says 148 on the server list so I'm guessing avg is 0.148s?
20:13 tenplus1     nite folks
20:13 Shara        That's my assumption anyway
20:13 Shara        And yes, 148 is what I was sitting at when using .15
20:14 Shara        so I guess not green
20:14 Shara        But.. I don't know anymore
20:14 Shara        I do feel a server runnign on .15 should be able to get green since it's completely playable
20:14 Shara        running*
20:17 shivajiva    hehehe ruben is testing me, have you seen areas code?
20:17 IhrFussel    0.15 - 0.20 should be green-yellow at least... < 0.15 green, 0.15 - 0.20 green-yellow, 0.21 - 0.274 yellow and > 0.275 red
20:17 Shara        Have looked at it a few times...
20:18 shivajiva    it's very objective contextually
20:18 Shara        IhrFussel: sofar is suggestnig 0.125 and lower for green, but I assume you mean server step now?
20:18 IhrFussel    I mean 0.201 - 0.274 yellow*
20:19 Shara        hmm
20:24 IhrFussel    I mean the value you see in the server list..like 148 for mine now
20:24 IhrFussel    I'm assuming those are the avg ms
20:25 Shara        I'm just wondering exactly how much difference remote media really makes
20:25 sofar        probably none
20:26 Shara        sofar: it's just that the problem seems to be on player join
20:26 sofar        hmmm
20:26 sofar        well then it's a legit problem
20:26 Shara        RC previously had the same issues DL seems to have, and remote media was one of the things that appears to have made a huge difference
20:27 sofar        what if 20 players all join at the same time?
20:27 Shara        DL is showing steadily decreasing max_lag until a player joins.. then it jumps straight back up again
20:27 sofar        I run mtmediasrv myself, I don't want the bandwidth to go through minetest
20:28 Shara        RC uses the TPS media server. Just wondering whether I will ask permission to add Dl's media there as well
20:28 sofar        I wish robbief would use mtmediasrv, using the index.mth method is just a waste of bandwidth
20:29 Shara        It's what existed at the time
20:29 Shara        He'd probably update
20:29 Shara        But he's kind of busy with non-MT things.
20:29 sofar        yes but mtmediasrv is like 6 months old by now, and tested in deployment for 3+ months
20:29 sofar        at least people whould understand why the index.mth method sucks
20:29 Shara        Poke him about it when you see him maybe? I don't know it well enough to mak e astrong case for switching
20:31 sofar        maybe I should just run a community media server
20:32 Shara        Could be nice. There's probably a few servers around that would get huge gains.
20:38 sofar        I'd want it on a different host than my server though
20:38 Shara        Well, yes. :)
20:38 sofar        right now it's on the same host as my server
20:38 Shara        WOuldn't make sense to hit your own performance
20:38 sofar        which is OK for my own server
20:39 Shara        You might want to set a limit on amount of media too. Some servers seem to have an insane amount.
20:39 sofar        with mtmediasrv, it's not a big deal
20:43 Shara        DL doesn't even have a huge amount of media. But I still don't really see what else it can be.
20:43 sofar        the problem is the connecting client without media is occupying cycles of the server thread
20:43 paramat      wut? someone joining causes 6s of lag? that's insane
20:43 sofar        and asking it for 500+ items by hash
20:44 IhrFussel    What is "huge"? My server has ~ 35 MB media
20:44 sofar        I've got 35mb+ as well
20:44 sofar        I'd say that's huge though, I have 10 audio tracks worth 17mb, and 36 skybox textures totalling another 15mb
20:44 sofar        if I take those out I'd be at 5-10mb
20:45 Shara        Each of my serevrs is around 10MB media I think
20:45 sofar        yeah, that sounds reasonable
20:45 Shara        RC does have a rather large space skybox for the space layer, and DL has mobs models, or they'd both be lower
20:45 sofar        not that 35mb isn't unreasonable
20:45 IhrFussel    Oops it's actually 45 MB xP
20:46 sofar        it's just that android clients will just time out
20:46 Shara        I've optimised quite aggressively though as well to push it down
20:46 sofar        (which is a win)
20:46 IhrFussel    If du -h is accurate 45M	.minetest/worlds/oldserver/worldmods
20:46 Shara        Well, not everything in mods directory will be media
20:48 IhrFussel    The biggest files in it are models
20:50 IhrFussel    Umm guys...the lag value in the server list doesn't really change
20:50 IhrFussel    My server is on 148 now since I restarted even though there were quite a few smaller lags already
20:51 Shara        DL moves about a bit.
20:51 Shara        RC only seems to move between 99 to 104... I think that's the range I saw between checking
20:52 IhrFussel    Then it must take a lot for it to change
20:54 sofar        that just means the value has consistency, which is good
20:54 sofar        if it would flap around all the time, that would be bad
20:54 Shara        t woul dhave been the first reset
20:55 Shara        eww, mis-tabbed :)
20:56 IhrFussel    sofar, does the server send that value or does the server list calculate it?
20:56 sofar        it's sent by your server
20:56 IhrFussel    Any clue in which file I can find it?
20:56 sofar        you could cheat it, but I know you have a conscience
20:57 sofar        src/server.cpp, I linked it earlier and in the github thread
20:57 IhrFussel    I don't want to cheat it, I want to see what causes this consistency
20:57 sofar        you can also not send it, btw
20:57 Shara        IhrFussel: check for m_lag in server.cpp
20:58 Shara        At least if I understood it right, based on where sofar said to look earlier, that will explain it
21:00 IhrFussel    Someone explain this math to me: is m_lag the result of all lag values added? m_lag += (m_lag > dtime ? -1 : 1) * dtime/100;
21:00 Shara        when m_lag > dtime it decreases, if I understand this
21:01 Shara        otherwise it goes up
21:03 nerzhul      Fixer, you problem seems to be related to windows only
21:03 nerzhul      it seems it's a WSAEFAULT
21:07 Fixer        okay
21:12 nerzhul      Fixer, i pushed another commit, if you can test it could be nice
21:12 Fixer        s u r e
21:13 nerzhul      i know it's long, but handling a shit OS is not easy, as it seems to do crap things :D
21:15 Fixer        c o m p i l i n g (please wait 6 minutes, i forgot to create yet another buildbot script)
21:15 Fixer        meanwhile -> beta 1.7.3
21:18 nerzhul      Fixer, git fetch && git checkout origin/network_asio instead of cloning
21:19 Fixer        nerzhul: i just do git pull and then compile it
21:19 Fixer        actually fuuuuuuuuuck
21:19 nerzhul      it works too
21:19 Fixer        ofc i can do simple mingw-make after that instead of running that script
21:20 Fixer        compiling done very soon
21:22 Fixer        linking that 160 mb exe is slow
21:22 Fixer        done, testing
21:22 nerzhul      debug powa :p
21:23 Fixer        :)
21:23 Fixer        (:
21:23 nerzhul      same thing it seems no ?
21:23 Fixer        connection lost
21:23 nerzhul      same log ?
21:23 Fixer        hmmmm, prelast line seems different
21:24 Fixer        i need to translate that crap somehow
21:24 nerzhul      interesting, show me
21:24 nerzhul      it's windows error
21:24 nerzhul      it seems your translation was not accurate i didn't found how to solve the previous error or somebody with it on google :p
21:24 Fixer        nerzhul: https://pastebin.com/raw/XsQJp6Zu
21:25 nerzhul      can you give me original message without modificaiton please ?
21:25 Fixer        no, it is in other language, good luck translating that
21:25 Fixer        message given by os
21:26 nerzhul      what is your language ?
21:26 Fixer        ukrainian
21:26 nerzhul      ouch exact it should be difficult :p
21:28 Fixer        let me try to find it
21:29 nerzhul      Fixer, i got the error
21:29 nerzhul      https://github.com/chriskohlhoff/asio/issues/86
21:29 Fixer        nerzhul: line #2 is WSAENOTSOCK 10038
21:29 nerzhul      it's this issue
21:29 nerzhul      IPv6 / IPv4 on UDP
21:29 nerzhul      on window
21:31 Fixer        error codes https://msdn.microsoft.com/en-us/library/windows/desktop/ms740668(v=vs.85).aspx
21:31 Fixer        yeah, and line #1 is very probably " "The system has detected an invalid pointer address in attempting to use a pointer argument in a call""
21:31 Fixer        WSAEFAULT 10014
21:40 nerzhul      Fixer, i got the solution
21:40 Fixer        okay
21:40 nerzhul      it's pushed, your turn now :D
21:41 nerzhul      i added a workaround by closing UDP socket and reopening it as v6 if remote_endpoint in TCP is V4
21:41 nerzhul      reopening as v4
21:41 Fixer        kicked in "mingw32-make package -j$(nproc)"
21:41 Fixer        should be much faster
21:41 Fixer        yeah it is faster
21:44 nerzhul      nice it works :D
21:44 Fixer        connected!
21:44 nerzhul      thanks winshit :p
21:45 Fixer        nerzhul: got crash
21:45 nerzhul      long but now there are useful logs on each endpoint to understand what happens
21:45 Fixer        lolol
21:45 nerzhul      Fixer, oh, give me the BT
21:45 nerzhul      but it's nice, we move forward :p
21:45 Fixer        need to rerun it under gdb
21:46 nerzhul      server crash wait a minute i'm doing a fix on it, it sems it's due to non MT users connecting on the server
21:46 Fixer        lol
21:46 Fixer        non MT users?
21:46 Fixer        chinese bots?
21:47 nerzhul      i don't know but the starting sequence trigger my huge packet detector
21:47 nerzhul      i should handle such cases and forbid them
21:47 nerzhul      server is up and i pushed some fixes
21:48 Fixer        the way falling nodes are moving is strange
21:48 Fixer        it is not smooth
21:50 nerzhul      yeah i think i should check if it's not send as reliable
21:50 Fixer        nerzhul: i have weird position problems
21:50 nerzhul      it seems you lag
21:51 Fixer        nerzhul: sometimes i'm transfered back to some position
21:51 Fixer        suddenly
21:51 nerzhul      yeah because you lag, server reset your position
21:51 nerzhul      (anticheat)
21:53 nerzhul      crash erf
21:53 nerzhul      seems the scanners are triggering some bugs i don't catch atm :p
21:53 Fixer        nerzhul: i don't have much problems with lag on any servers, this position reset is huge
21:54 Fixer        nerzhul: packets flow and then stop and then after some seconds LOTS arrive
21:54 nerzhul      yeah we should anaylyze that but i'm tired
21:55 nerzhul      at least we fixed your connection problem on windows, and for all windows
21:55 nerzhul      maybe jas_ will not have problem now :)
21:55 Fixer        is there really any benefits of this?
21:55 Fixer        asio thing
21:55 nerzhul      code maintenance, network performance and handling on laggy clients
21:56 Fixer        nerzhul: crash on joining singleplayer
21:56 nerzhul      current network doesn't resist well when a slow client connects
21:56 Fixer        nerzhul: backtrace follows
21:56 nerzhul      i will go to bed, give me backtraces tomorrow :)
21:57 Fixer        nerzhul: one sec
21:57 rubenwardy   any recommendations for good meta privileges mods? For example, I can make a meta privilege called "moderator" which gives certain permissions, and takes them away when revoked
21:58 Fixer        nerzhul: https://pastebin.com/raw/mas1HY15
21:59 Aerozoic     rubenwardy, i'm not sure i follow, got a better example?
21:59 Fixer        nerzhul: and winsock gave different errors
21:59 rubenwardy   so I can do   /grant playername moderator, and it gives them ban, kick, basic_privs but not a priv called moderator
21:59 rubenwardy   as there is no priv called moderator
22:00 rubenwardy   like how all is a meta privilege that grants or revokes all privileges
22:00 rubenwardy   I'm sure I've seen a mod like this
22:00 Aerozoic     ohhhh ok
22:01 nerzhul      Fixer, disconnection issue
22:01 nerzhul      does MT told you in console why ?
22:01 Fixer        nerzhul: there were two errors
22:02 Fixer        nerzhul: 1) invalid name of the network
22:02 rubenwardy   doesn't have to work like that
22:02 Fixer        nerzhul: 2) input-output operation terminated because stop of flow of commands or on interrupt
22:03 rubenwardy   any way of letting me define groups of players with certain privileges would be nice
22:03 Fixer        or interrupted by application (or smth)
22:03 rubenwardy   like how IRC's access lists does it
22:03 Fixer        can't find suitable winsock error
22:03 Fixer        probably #1 is WSAENETUNREACH
22:04 Fixer        10051
22:04 nerzhul      i push a test to ensure socket is opened before closing it, just test it and tell me tomorrow, i'm very very tired and need to sleep
22:04 nerzhul      yeah WSAENETUNREACH is the problem
22:04 nerzhul      without doubt
22:04 nerzhul      the crash is a bad disconnection handling
22:04 Fixer        nerzhul: i've tested on outdated client btw
22:04 Fixer        i will check newest
22:05 Fixer        tested on d865
22:05 nerzhul      not so old :p
22:07 Fixer        nerzhul: if you wait just a little, i will check
22:07 Fixer        like 1-2 minutes
22:07 nerzhul      okay, but it's the last time :p
22:07 Fixer        creating package
22:08 Fixer        nerzhul: yeah, crash
22:09 Fixer        3rd error with "remote host closed connection"
22:10 Fixer        everything else looks the same
22:10 Fixer        same interlocked thing in backtrace
22:10 nerzhul      we will look tomorrow
22:10 Fixer        /quit
22:11 nerzhul      other thing i forgot to tell, TCP usage permits to reduce latency as stream will be sent as biggest packets on network, MT uses 500 , but with a TCP stream it's dynamicly adjusted and generally set to 1400-1480
22:11 nerzhul      for reliable packets it's faster
22:13 nerzhul      ty for your time
22:14 Fixer        np
22:17 Shara        So at 21 players DL is reporting in the same range for max_lag as when there's only 5 players...  issue being with on join seems more and more likely.
22:18 Aerozoic     rubenwardy, i could see that useful for owners. I could create groups like moderator and admin with the appropriate privs for each.
22:19 Aerozoic     groups = meta priv
22:20 Aerozoic     btw rubenwardy, any chance that privilegeareas mod will ever be fixed?
22:21 rubenwardy   people still use that? :D
22:21 Aerozoic     I really like the idea of it, i installed it for my server then discovered that critical issue.
22:22 Shara        I always liked the idea of it.
22:23 rubenwardy   well, I'll add it to my to do list
22:23 Aerozoic     :D
22:23 rubenwardy   keep in mind that it currently has just under 100 entries on it
22:23 Fixer        Shara: just curious, can you join server and type F6 and look at first SEnv step?
22:23 Aerozoic     lol, yeah i figured
22:23 Fixer        Shara: your server
22:24 Shara        Sure.
22:24 Aerozoic     rubenwardy, but shouldn't it get bumped up just a little since i reported the issue months ago?   :)
22:24 Shara        Tell me what I'm looking for?
22:25 Shara        Hmm, don't see SEnv. Just CEnv?
22:26 Fixer        type F6 more
22:26 Shara        I did, but not seeing it
22:26 Shara        Three pages total
22:26 Fixer        SEnv should be there
22:27 Shara        First page has things starting CEnv and CM
22:27 Fixer        page 2
22:27 Fixer        SEnv step
22:27 Shara        Second has more CM things... Client .., Elasped time and FPS
22:28 Shara        Last page has mesh update, meshgen, rendering of clouds, sky, and collision things
22:28 Fixer        okay
22:28 Fixer        nevermind then
22:28 Fixer        ty for checking
22:29 Shara        No problem. More than happy to look into anything if people have ideas
22:34 Shara        Maploading seems quite fast, but lag keeps reporting high.
22:35 Shara        I've asked a few players if they feel much lag while playing, and they seem to think it's fine
22:35 Shara        I just don't get it
22:35 Shara        Even chests are responding instantly at the moment, which is usually where I notice a delay if I'm going to
22:38 Fixer        Shara: try IhrFusse ls tool
22:39 Calinou      I pushed the README rewrite thing to Map Tools too :P
22:39 Fixer        Calinou: rewrite all the readmes
22:39 Calinou      "README as a Service"
22:39 rubenwardy   I just snipped someone from ~50m, and the idiot just kept looking around and didn't hide so I killed them
22:39 rubenwardy   silly mobile player
22:39 Calinou      rubenwardy: mobile players are just cannonfodder in CTF :P
22:39 rubenwardy   yeah
22:40 Shara        I'm wondering if in general the lag is actually normal and it's just the joins keep pushing ip up with enough quick spikes that it keeps reporting high
22:40 Shara        it*
22:40 rubenwardy   the last match was quite fun as my flag kept getting stolen whilst I was carrying the enemy's back
22:40 rubenwardy   so had to keep intercepting and killing them
22:40 Shara        But that still means the joins are a problem.... maybe will try FUssel's mod though
22:40 Fixer        Shara: take fussels lag tool, set fast report interval and observe the lag when players join
22:41 Fixer        Shara: just chainge two numbers in it from 60 sec to smth like 1 sec
22:41 Fixer        change
22:42 Shara        Yes.. but not tonight.
22:42 IhrFussel    If you set it to 1 sec it will simply tell you the realtime dtime
22:43 Fixer        yeah
22:43 paramat      when someone joins do players actually experience server lagging for 6s? because that would be a very annoying lag
22:43 IhrFussel    Well the realtime dtime average of the recent second*
22:43 Shara        Need to spend a little time staring at sfinv before I'm too tired to
22:44 Shara        paramat: no, but max lag reported goes up by 1 to 2
22:44 Fixer        paramat: maybe it is that thing sorcerykid reported about larghe auth.txt?
22:44 Shara        Then it starts to fall again, until another player joins
22:44 Shara        And they are joining often enough that the max lag never seems to get to less than 3 (and rarely that given how busy the server can be)
22:44 IhrFussel    Shara, the auth.txt gets rewritten in on_joinplayer() yes
22:45 Shara        I might need to look at reducing auth file size as well
22:45 IhrFussel    To update the last login timestamp
22:45 Shara        Good idea
22:45 paramat      maybe i misundertand and 'max lag' is not actually the experienced server lag
22:45 IhrFussel    Also the auth file is rewritten whenever you change privs
22:45 Shara        I pruned RC's auth file from 34MB to less than 3MB today...
22:45 Fixer        paramat: max_lag is moving exponential average (for pretty long time) so it kinda useless for understanding current lag
22:46 Shara        For me, max lag just tells you if it's getting better or worse
22:46 Shara        So if it's gone up shareply, you know something bad happened
22:46 Shara        If it's decreasing, things are good
22:46 Fixer        Shara: lag over ... say 10 minutes
22:46 Shara        sharply*
22:47 Fixer        Shara: you can have lag 10, yet it is not 10 anymore, but 0.020
22:47 Fixer        thats why I ask about curr_lag in /status
22:47 Fixer        to stop this misunderstand of statistics
22:47 Fixer        IhrFussel: you should post your lag mod on github and forums, it is useful
22:47 IhrFussel    avg_lag should be about what my /lag reports
22:48 Shara        Yea, it would be nice to have a more immediately value
22:48 Shara        immediate*
22:48 paramat      here's sorcerykid's issue https://github.com/minetest/minetest/issues/5334 any relevance?
22:48 Fixer        possibly, Shara needs to install that mod and do per second measurement and correlate it with people joins
22:49 Shara        Fixer: I know it correlates
22:49 Shara        paramat: yea, might be related
22:49 Shara        DL is definitely high traffic with lots of joins
22:49 Fixer        also, look up size of your auth.txt
22:49 IhrFussel    Shara, my mod tells you the max_lag of the recent second so you can exactly see if the lag happened at the moment of the joining 100%
22:50 IhrFussel    You need to set 60 to 1 of course
22:50 Shara        Fussel, yea, but like I said, not touching mods there again tonight
22:50 Shara        Fixer: I am pretty certain auth will be an issue too now it's been mentioned
22:51 Shara        At what size would you say auth is too big?
22:51 IhrFussel    Fixer, will make it standalone later and post on forums...not sure how to post something on github yet
22:52 IhrFussel    I will also add a simple var at the beginning of the file to set seconds
22:52 Shara        Fussel, happy to help you with getting it on github if you need
22:52 Shara        This would be a very nice mod for people to have access to
22:53 IhrFussel    Is it just 1 command to upload to github?
22:54 IhrFussel    That would be too easy I think x)
22:54 Shara        You need to start a repo first, then add the files, then commit, then push
22:54 rubenwardy   don't upload zips
22:55 Shara        That too :)
22:55 rubenwardy   upload the mod files with no directory. So init.lua should be in the root of the repo
22:55 rubenwardy   I suggest trying Github Desktop
22:55 paramat      reopened the issue
22:56 * Shara      is having seriously thoughts about trying shivajiva sauth mod there
22:56 Shara        's*
22:56 rubenwardy   y?
22:56 Shara        The auth file is just short of 40MB
22:57 Shara        RC's auth file corrupted last night at slightly less than that size
22:57 rubenwardy   so you'd like to try it?
22:57 rubenwardy   ah
22:57 rubenwardy   I think you meant you were having second thoughts about trying it
22:57 Shara        rubenwardy: just wondering if it would help reduce problems from player joins, since that's what seems to be causing my issues on DL
22:58 Shara        And it's been suggested writing to auth.txt could be at least part of that
22:58 rubenwardy   it should do as sqlite should be able to handle more data than a file
22:58 rubenwardy   otherwise why would it be used?
22:58 Shara        That's what I'm thinking
22:58 Shara        auth being an issue could also explain why this has got worse over time
22:59 Jordach      >goes to frontier expo
23:00 Jordach      >gets full on MIT talk about biological life
23:02 Calinou      frontier expo?
23:02 Fixer        Shara: lol,mb
23:02 Fixer        40mb*
23:02 Fixer        impressive
23:02 Calinou      rubenwardy: oh, it's nice, you have an official CTF server again
23:02 Calinou      what's the server specs?
23:02 Calinou      it has 19/21 players, probably due to connecting players, heh
23:03 Calinou      I like how you can call a server "Capture the flag" and it instantly gets tons of players, 24/7
23:03 rubenwardy   ha
23:04 rubenwardy   I think it's an i5 with 8GB ram and a 1TB hard drive
23:04 Fixer        Shara: quite strange to see that huge auth.txt on servers, there are that much players? or there is this stuuupid shit with those mobile joining with different nickname each time?
23:04 rubenwardy   may be wrong on the CPU
23:04 rubenwardy   ah, i3
23:04 rubenwardy   Intel(R) Core(TM) i3-2130 CPU @ 3.40GHz
23:04 Shara        Fixer:  DL is very busy
23:04 Fixer        rubenwardy: my CPU ^_^
23:04 Shara        Like.. quite often constant joins
23:04 rubenwardy   CTF doesn't need a SSD as the hard drive use is non-existent
23:05 Fixer        Shara: i have an idea
23:06 Fixer        Shara: some mod that calculates players statistics, how often they join, for how long they play, avg play time between rejoins etc, number of player who no longer play, etc
23:06 paramat      wow 40MB
23:06 rubenwardy   Calinou, there are 2 other servers - OldCoder CTF and ESP CTF
23:06 rubenwardy   but they are consistently at 0-2 players
23:06 rubenwardy   not sure why
23:06 rubenwardy   lag maybe?
23:06 paramat      hehe
23:06 Shara        I should check what's happening with ESP.
23:06 Shara        It used to be crazy busy
23:07 Fixer        Shara: they can't just constantly join, they fumble for like few minutes and probably quit? otherwise you will have full server
23:07 Shara        Server is pretty full
23:07 Shara        But yes, obviously they leave too
23:07 Fixer        hometown had 67 or smth players recently o_O
23:07 rubenwardy   > uses server password on PC to sudo
23:07 rubenwardy   urgh
23:08 Fixer        sudo is life
23:08 Shara        Thing that makes me think it must be on join is that the reported lag is in a constant range no matter the actual numbe rof players. Justs eems to correlate with joins. Others have confirmed now.
23:10 Fixer        rubenwardy: for some reason thinked about that South Park episode with hand written porn, so good, need to rewatch
23:10 Fixer        thats why you have local copy of everything
23:10 Fixer        not really everything, most important parts
23:11 rubenwardy   Calinou, 0.05s dedicated_server_step, not visible jitter :)
23:11 rubenwardy   may try reducing to 0.02-0.03
23:11 paramat      how to be popular: try to be MC. i suspect the popularity of multicraft is down to this, there will obviously be a huge number of people who want a 'free MC'
23:11 Fixer        paramat: and you ask me not to measure to mc standard, mc is standard, and people will ask this
23:12 Fixer        people expect basic mc features
23:12 Fixer        look for example inventory management
23:12 Fixer        minetest is light years behind in this and has like shift-click as top feature, thats mc beta like
23:12 Fixer        entity movement by liquids, other interactivity
23:13 Fixer        maybe we need to include mineclone2
23:13 Fixer        to minetest
23:13 Fixer        and rename it to smth better
23:13 Fixer        mineclone2 to minesomething
23:13 Fixer        obvious to minecrafters
23:13 Fixer        and implement mc like stuff there
23:13 Calinou      rubenwardy: 0.05 is good enough, I guess, 0.03333 (30 Hz) is probably too much right now
23:14 rubenwardy   I suspect it will cause issues
23:14 Fixer        include in minetest at least: 0) mtg 1) tutorial from wuzzy 2) mineclone2 3) boxes from sofar 4) other?
23:14 rubenwardy   especially for mobile clients
23:14 Fixer        rubenwardy: are you on server right now?
23:15 rubenwardy   not MC2, there's going to tm/patent issues most likely
23:15 sofar        insidethebox is not going to be offline
23:15 rubenwardy   I can be in 10, need to brb
23:15 sofar        not any time soon, and I don't see the point
23:15 paramat      "and you ask me not to measure to mc standard" that's irrelevant
23:15 Fixer        rubenwardy: tm/patent issues? o_O rename all stuff?
23:16 rubenwardy   urgh
23:16 Fixer        paramat: mc has lots of good features, naturally people want that backported into mt
23:16 rubenwardy   I think I'm going to have to remove vote on CTF
23:16 rubenwardy   or limit it heavily
23:16 Fixer        lol, full CTF server
23:16 Calinou      rubenwardy: I'm not sure if Microsoft/Mojang have any *patents* on Minecraft
23:16 Calinou      there would be copyright and trademark issues, though
23:16 Fixer        patented snow blocks? wth
23:17 Fixer        shulker boxes? just rename it
23:17 rubenwardy   lol, the server is 1 in, 1 out
23:18 Fixer        i don't see much of a problem there, sensitive stuff can be renamed
23:18 paramat      it's unreasonble for people to expect MT to be an exact MC clone
23:18 Shara        rubenwardy: told you vote is annoying :)
23:19 Fixer        not exact, but a lot of mc features are cool/nice in gameplay and people like it, including those may add more popularity to minetest game
23:19 ThomasMonroe is nore on?
23:19 Fixer        or some obvious stuff
23:19 Fixer        minetest game has weapons
23:20 Fixer        where is weapons you have shields (bingo
23:20 Fixer        )
23:20 Fixer        you have pvp = knockback
23:20 Fixer        and other pvp effects
23:20 Fixer        you have farming = suggests smth
23:20 paramat      yes i did think that a mineclone game would pull in players, not sure if wuzzy would allow including it
23:20 Fixer        etc
23:21 Fixer        lol, wuzzy very probably wants it
23:21 ThomasMonroe hey Fixer Caleb had a really cool idea for tool
23:21 ThomasMonroe tools*
23:21 Fixer        iirc, he had frustration on why people work on mtg only, but not on others too
23:22 Fixer        ThomasMonroe: what tools?
23:22 ThomasMonroe have different types of workshops, ie a grinding wheel for increasing dig time
23:22 ThomasMonroe er decreasing
23:22 Fixer        work he done on mc2 is quite impressive
23:22 ThomasMonroe or a CNC for making the tool lighter for faster swing time
23:23 paramat      but yes i see possible issues with mineclone being so similar to MC
23:23 Fixer        maybe thats already implemented, those drills, but i have not used it
23:24 Fixer        to me minetest_game is still basicly early minecraft beta clone, with no armour and no automation, even some textures were well mc betish
23:24 Fixer        i can feel cl55 used mc beta as a kind of benchmark
23:25 Fixer        minetest_game has much more blocks
23:25 Fixer        they are nicer too
23:25 Fixer        variety of mapgens, that btw need richer biomes, they are pretty simple right now
23:26 paramat      ~tell twoelk please could you keep your leave message very short, it's spammy and is not meant to be used as a forum signature =)
23:26 ShadowBot    paramat: O.K.
23:28 paramat      MTG is meant to be simple, very rich biomes are for other subgames or mods. however we are working on more flora
23:28 * Shara      mutters something about underground things and then hides
23:29 paramat      there's not much point listing everything MC has that MT doesn't, they're differnet games, and most of what you suggest is already intended, it's just down to dev time and priorities
23:30 rubenwardy   I think we need more texturers
23:30 rubenwardy   for me, most of the time adding something is in making new textures
23:30 paramat      yes underground decorations is within my next 2 tasks
23:30 rubenwardy   how about making issues for each thing we want to add, and asking for support?
23:30 rubenwardy   Fixer, joining ctf now
23:31 rubenwardy   if you tell me when you're going to join, I can let you in
23:46 Fixer        rubenwardy: unf-ly have to go, bye
23:46 rubenwardy   o/
23:53 IhrFussel    max_lag just spiked to 9.2 secs with 4 players...something I didn't see before
23:54 IhrFussel    Not caused by any command since I log those