Time  Nick        Message
02:59 Emerald2    What the heck. Going near a uranium block makes you keep taking damage even after you leave the area?
03:09 Emerald2    Fun new things in technic. :p
04:13 diemartin   Magic.
04:15 VanessaE    witches.
04:15 VanessaE    or wormholes.
04:30 Emerald2    I had to relog from game to stop dying from being radioactive. XD
04:30 Emerald2    I didn't know I was about to walk up to a block of 3.5% uranium. :P
04:31 VanessaE    derp :)
04:49 diemartin   !rainbow ?????
04:49 MinetestBot ?????
04:51 diemartin   ? MinetestBot
05:03 sofar       Emerald2: you're just looking positively radiating today
05:03 sofar       (I can see your tail in the dark)
05:03 VanessaE    heh
05:07 Emerald2    lol sofar
05:07 Emerald2    My... tail?
05:08 sofar       radiation, mutation
05:09 VanessaE    yeah, you can join the X-men now :)
05:10 Emerald2    Worst super power ever.
09:32 Krock       o/ u_i finally got an API documentation
09:38 Emerald2    G'day Krock. o/
09:38 Krock       o/ Emerald2
11:36 IhrFussel   Does get_objects_inside_radius() always take about the same time to execute? It looks like it... no matter if I specify as radius 20, 10 or 5 ... the profiler tells me it takes about the same time
11:36 IhrFussel   I guess it would make sense cause the engine has to loop through all active mapblocks in any case
11:37 rubenwardy  no
11:37 rubenwardy  the whole point of having mapblocks is so that you don't need to
11:37 rubenwardy  the engine will only iterate through mapblocks within that radius
11:38 IhrFussel   I mean the engine has to look through all loaded entities to be able to list and push the ones that are inside the radius
11:38 rubenwardy  actually, that PR was reverted
11:38 rubenwardy  so it does have to look through all entities
11:38 rubenwardy  which is stupid
11:39 IhrFussel   Someone tried to change that in 5.0.0?
11:39 IhrFussel   Or is it older?
11:40 Krock       was a PR for 0.5.0, early development of 0.4.17 I believe
11:41 IhrFussel   Why was it reverted? Unexpected behavior?
11:41 rubenwardy  bugs
12:48 IhrFussel   Uhh...is this supposed to be sarcasm? https://forum.minetest.net/viewtopic.php?f=3&t=22458
12:50 rubenwardy  yeah
12:52 IhrFussel   The title is clickbait then
12:52 IhrFussel   Should've written something like 'What MT 6 could be like'
12:54 rubenwardy  I better move it to News
12:55 Krock       lol
12:55 Krock       offtopic storytime
12:58 IcyDiamond  LOL
13:33 IhrFussel   Festus1965 claimed that any server with 40+ players is unplayable... well it is possible to keep it playable by optimizing any heavy code
13:33 IhrFussel   Right now my max_lag is 0.35 s with 46 players
13:34 IhrFussel   36*
13:38 IhrFussel   Only 4 of them are idling too... so 32 actively playing people
13:42 IhrFussel   He now lowered his dedicated_server_step even more
13:42 IhrFussel   'dedicated_server_step = 0.016'
13:42 IhrFussel   I swear if he has ANY mods that run on every globalstep...
13:46 IhrFussel   The default of 0.09 is not for multiplayer servers btw ... cause I#m pretty sure there are engine internal things that run on every step and with 40-50 players this gets repeated too often... or does every engine function also use a timer?
13:49 IhrFussel   I'm talking about things like 'check if mapblocks need to be sent/updated to clients' ... AFAIK this check happens every step right?
13:56 Krock       0.09 is fine for 10-ish players. 0.2 might be a better choice for a large userbase like on your server
13:56 IhrFussel   I have it at 0.15 and it works well with ~ 40
13:58 Krock       and mapblock sending is async anyway
13:58 IhrFussel   Guys we need to find the source of the 'map lag' problems ... They happen more and more on my server whenever too many users with a certain app are connected ... I wasn't able to find out which app it is yet but I'm afraid that the name of the app won't help us at all cause it's likely closed sourced
13:59 IhrFussel   Either the map loads very very slowly or it stops loading completely
14:00 IhrFussel   Until a certain app user leaves... but if only 1 leaves the map starts to load VERY SLOW again... if more than 1 with the app leave it loads afster again
14:00 IhrFussel   It's like each app user adds some overhead to the map
14:01 IhrFussel   I mean adds more overhead than normal
14:01 IhrFussel   I set my 'max block sends per client' to 10
14:02 IhrFussel   max block send distance is at 5
14:02 IhrFussel   So pretty low
14:08 Krock       well, block sending is per-player
14:08 Krock       the only possible explanation is that they would discard all received mapblocks
14:09 Krock       because the importance of sending the mapblocks is distance-based, not just time-based
14:09 Krock       -just
15:10 Fixer       hey,littledrums
16:17 Krock       Minetest idea: wireless data transfer, using raycast to check whether the view is given
17:42 IhrFussel   Krock, does your explanation also explain why entities still load even in 'unloaded' mapblocks?
17:43 IhrFussel   All entities still load and function as expected when the map stalls
17:44 IhrFussel   And when I take something out of a chest during the stall it appears on my client as if I doubled the stack
17:44 IhrFussel   Meaning it will be in my inventory and still in the chest
17:44 IhrFussel   When the stalling is over the stack in the chest then disappears
17:45 sofar       that's client side prediction
17:47 IhrFussel   Makes sense... still, mobs still load in the 'void' and walk around...they are clearly working
17:47 IhrFussel   So it must be something that doesn't affect entities at all
17:49 IhrFussel   Entities are stored inside the mapblock though right? So those bad clients must trigger something that happens only after the server extracts the entities
17:49 calcul0n_   this is "normal" behaviour of minetest, i see this all the time when the server is lagging
17:49 calcul0n_   don't know much about that but it looks like entities and mapblocks handling are totally unrelated
17:50 sofar       why are you assuming "bad clients"
17:50 IhrFussel   Cause it only happens with randomly generated names mostly and only when there are LOTS of them connected
17:51 IhrFussel   I can see exactly WHO caused it cause as soon as they leave the map works just works again
17:51 IhrFussel   the map just works again*
17:51 sofar       so you're just guessing
17:52 sofar       have you ever heard of confirmation bias?
17:52 IhrFussel   It's like this -> Less than 5 mobile users connected = no map lag ... 10+ connected map gets slower and at 15 or 20 (approx.) the map stalls completely until 1 of them leaves
17:53 IhrFussel   These are no guesses! It is not caused by the official Android app
17:53 sofar       do you have evidence that this doesn't happen when there are equal amount non-mobile players?
17:54 sofar       or are you only looking at your server when there are high amount of mobile players?
17:54 IhrFussel   I have evidence that it suddenly works again as oon as a MOBILE app user leaves
17:54 IhrFussel   Are you telling me it's normal that the map doesn't load for 1 minute based on amount of players?
17:55 sofar       despite of that, I see no reason to suggest that this isn't an artifact of the engine design and there could be a huge bug
17:55 sofar       but it would be interesting to do a statistical analysis on the traffic to/from those clients and see if they are somehow making the server behave differently
17:55 IhrFussel   So far it only happens when my server is full of people AND most of them are app users
17:55 sofar       I assume this is a 0.4.x server, right?
17:56 IhrFussel   0.4.17.1 yes ... we would need a 5.0.0 server full of mobile players to test if it happens there too
17:57 IhrFussel   But I'm sure that it only happens with mobile users...cause 0.4.17.1 CTF also had this issue sometimes
17:57 IhrFussel   It was very rare cause CTF disallows these names... so only when someone installed such a 'bad app' and changed their username in the app they could connect and cause this
17:58 sofar       reminds me I should further brush up the wireshark plugin
17:59 IhrFussel   Something else I observed: It looks like these app users can still walk around the map while my client doesn't receive any parts
17:59 IhrFussel   They were able to walk away from me into the void
18:00 IhrFussel   Maybe their app code tries to prioritize the mapblock sending for them somehow and messes up the server this way...I don't know much about the map sending c++ code
18:01 sofar       there's a queue per player
18:01 sofar       you can limit the queue size, even
18:01 IhrFussel   Isn't that the 'max block sends per client' ?
18:01 sofar       yup
18:01 IhrFussel   That is set to 10
18:02 sofar       super low?
18:02 IhrFussel   Default is 12 I think
18:02 sofar       max_block_send_distance
18:02 IhrFussel   5
18:02 sofar       you can lower that one
18:02 IhrFussel   It is 5... should be low enough
18:03 IhrFussel   You think it's possible the app demands more than the server wants to send and stalls it this way?
18:03 sofar       max_simultaneous_block_sends_per_client
18:03 sofar       you can lower that
18:03 sofar       max_packets_per_iteration
18:04 sofar       that's something you can lower as well
18:04 rubenwardy  I suspect that some clients are filling up the quota for map loading, starving the other users
18:04 sofar       well, it would point to a DB bottlenect
18:04 sofar       neck*
18:04 IhrFussel   I thought there was no hard coded 'total' limit anymore
18:05 IhrFussel   I read somewhere that the setting is not used anymore cause it basically is the result of blocks per client
18:06 sofar       block_send_optimize_distance may influence bahavior too
18:07 Krock       IhrFussel: entity sending and mapblock sending are separate
18:08 Krock       rubenwardy: same as my thought but it's limited per-player
18:08 sofar       Krock: we could make the code send blocks around entities automatically
18:08 rubenwardy  isn't there a total limit too, calculated from the per-player limit?
18:08 Krock       yes sure but that doesn't solve the problem here
18:08 sofar       alternatively, we could *not* send entities in unsent mapblocks
18:08 Krock       rubenwardy: that too, on top of the per-player limit
18:09 Krock       it's limited by client and by async server step
18:14 IhrFussel   Krock, then I guess entities have no limit
18:15 IhrFussel   That's why every client can still receive them no matter how hard the server struggles with the map
18:16 IhrFussel   'I suspect that some clients are filling up the quota for map loading' so the client can ignore the set limits server side under certain cicumstances?
18:20 IhrFussel   Should the mapblocks be visible as pps in 'bmon' ?? I mean if the server sends lots of mapblocks for a while will those pps go up?
19:16 IhrFussel   Do 3K packets sound normal with 20 users?
19:16 IhrFussel   packets per second*
19:19 IhrFussel   Not linked to pps ... map stalls even with 1.5K pps
19:24 IhrFussel   Most users who cause this use an app with protocol version 27
19:32 Krock       strict protocol version = true
19:48 IhrFussel   But we need to fix this...else the same apps might cause map stalling when they upgrade to 5.X
21:32 VanessaE    Just spamming the channel....  <beg> https://www.gofundme.com/f/5pqdb-help-me-move? </beg> :)
22:22 rubenwardy  Would anyone be interested in hosting CDB mirrors?
22:25 VanessaE    If I didn't have this moving thing looming over my head, I'd offer.  I have about 800 GB of available space on daconcepts.
22:26 VanessaE    (as long as you don't mind ssh'ing in and maintaining it ;) )
22:27 rubenwardy  I'll check how much bandwidth I have available. The idea is that it would make backing up the uploads much nicer, and hopefully have localised downloads
22:27 rubenwardy  I wouldn't be free to do it for many weeks anyway
22:29 VanessaE    well I have that 800+ GB sitting behind a decent 100 Mbps pipe out of Canada, so just let me know.
22:29 rubenwardy  oh nice, there is no bandwidth/transfer quota. So that lets the pressure off for now
22:30 VanessaE    same here afaik
22:30 VanessaE    plus, DDoS protection comes standard.
22:30 rubenwardy  don't we both have the same model?
22:30 rubenwardy  Kimsufi i3 whatever
22:30 VanessaE    we might?  Kimsufi KS-7 here
22:31 VanessaE    yep, the i3-2130 package
22:31 rubenwardy  Kimsufi KS-2C - 8G i3-2130 1x1TB Server
22:32 VanessaE    the KS-7 package has 2TB storage... hard to imagine I already soaked up half of that with my server stuff huh :)
22:32 VanessaE    tell you what though.. the SSD's in the old SYS server had me spoiled.
22:33 VanessaE    my morning mapper script takes 7 hours to finish now.  used to take about 2.
22:34 rubenwardy  "morning mapper script"
22:34 rubenwardy  did you mean *minutes*?
22:34 VanessaE    hours.
22:34 rubenwardy  whaaaaaaat
22:34 rubenwardy  how big is your map?
22:34 rubenwardy  which mapper do you use?
22:34 VanessaE    there are 6 maps to be processed, two of which are in the tens of gigs range
22:34 rubenwardy  are you tiling it?
22:35 rubenwardy  oh dear
22:35 VanessaE    tiling with leaftest, and using minetestmapper
22:35 VanessaE    (i.e. the official version)
22:35 VanessaE    lesse...
22:35 rubenwardy  ah yes
22:36 VanessaE    37GB for dreambuilder creative
22:36 rubenwardy  leaftest takes a while
22:36 VanessaE    20 GB for dreambuilder survival
22:36 rubenwardy  https://rubenwardy.com/redcrab/downloads/   <-- I ran it here on a *100MB* file, and it still took a while
22:36 VanessaE    the others are under 10 GB, smallest is skyblock, at just 13 MB.
22:37 VanessaE    hey, I can see my house from here! :P
22:37 VanessaE    (just north of the water pyramid)
22:37 rubenwardy  leaftest outputs JPGs for the closest zoom, and it stresses me out
22:37 rubenwardy  so many artifacts
22:39 VanessaE    mine doesn't?
22:39 VanessaE    it outputs PNG all the way through
22:39 rubenwardy  huh
22:39 rubenwardy  I use a fork of minetestmapper, maybe that's why
22:40 rubenwardy  one that does the isometric you see in the header on thatpage
22:43 VanessaE    that's onomatopoeia isn't it?
22:44 VanessaE    or just some new feature?
22:44 rubenwardy  it's another mapper
22:45 VanessaE    oh ok
22:47 VanessaE    btw I think you have the spawnpoint marked wrong on that map.  I want to say it's the intersection to the right of the white-bordered building that's to the right of the orange/white checkered building...?
22:47 VanessaE    been too long since I've been there.
22:48 VanessaE    no, that's not right either...
22:48 VanessaE    (I'm going from memory, the path I recall taking to get from spawn to my house :) )
22:51 rubenwardy  the leaflet co-ordinate system doesn't work
22:51 rubenwardy  or maybe it's just that fork again
22:51 rubenwardy  it's weird
22:52 VanessaE    you're welcome to fiddle around with my mapper script, maybe it'll help?
22:53 VanessaE    https://daconcepts.com/vanessa/hobbies/minetest/Server-scripts/update-server-maps-leaftest.sh
22:55 VanessaE    one thing that seems obvious to me though is that leaftest or the mapper or *some*thing, reads through the entire database when generating a map, even if you've only asked for a small portion of it
22:56 VanessaE    because as the map database gets bigger, the mapping time increases, even if the area being mapped doesn't have any large changes