View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0004414 | Spring engine | General | public | 2014-06-04 09:53 | 2017-01-07 14:26 |
| Reporter | Anarchid | Assigned To | Kloot | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | closed | Resolution | no change required | ||
| Product Version | 97.0.1+git | ||||
| Summary | 0004414: 97.0.1-3 experiences significant performance loss on player roster changes | ||||
| Description | Assuming sufficiently large player count (say 10 people total), joining or removing a "player" - including spectators - seems to cause a massive temporary performance decrease for everyone involved in the room. This was repeatedly and reproducibly observed on Springie hosts running ZK with Spring 97.0.1-3. Note that since this also happens with observers, the issue is not likely related to any handling of units or associated lua. It was not possible to test prior versions yet with sufficient player counts that would make this issue easily reproducible. Same applied to testing with examplemod or BA. For that reason, the ticket could be considered incomplete. I will fill in the remaining data as it is obtained. For the reason of incompleteness, i've filed it as "minor" severity. | ||||
| Steps To Reproduce | 1) Join a ZK teamgame. 2) ??? 3) Suffer. | ||||
| Additional Information | Are there other ways to profile this besides the ones i outlined in description? | ||||
| Tags | No tags attached. | ||||
| Attached Files | |||||
| Checked infolog.txt for Errors | |||||
|
|
This is true, I played this sandbox game http://zero-k.info/Battles/Detail/265855 with only 3 player, but when another person join.... it stutter/pauses/lag a lot for few second. Its like player joining create an infinite loop. Its like some nuke exploding ingame. |
|
|
copy/pasted as request of Anarchid [11:03] <bluestone_irc> "Note that since this also happens with observers, the issue is not likely related to any handling of units or associated lua. " > use /debug while it happens to check this [11:04] <xponen> *foreheadpalm* I forgot! [11:05] <Anarchid> just don't try foreheadnapalm [11:05] <Anarchid> bluestone_irc i tried that and saw nothing i could process into useful conclusions [11:05] <bluestone_irc> also, check if it reproduces in demo [11:05] <Anarchid> oh that's a good idea [11:05] <Anarchid> i've actually added a question for "how else to profile this" in ticket, can you post it thar? [11:07] <bluestone_irc> maybe even better way is to run gadget/widget profilers while it happens |
|
|
"seems to cause a massive temporary performance decrease for everyone involved in the room." what does that mean? high cpu usage? game slows down but fps keeps the same? fps drops? |
|
|
FPS significantly drops for everyone, simframes seem to take longer to generate. |
|
|
does the fps increase after some time again? how long is the "drop"? |
|
|
It seems to generally rebound after a few minutes. |
|
|
do we have a replay where this happened? edit: forget it. did someone test if replaying the demo causes the fps drop, too? |
|
|
I did a quick test with the replay I've just attached (Anarchid gave me it). Players call "laaggg" at about 8 minutes in. In the replay it correlates precisely to a prolonged spike (maybe 30 sec) of abnormally high usage for the "Tactical Unit AI" synced gadget. I didn't test anything else and I don't know much about this report so please don't take my result as a "there is nothing else here" without asking someone who actually did investigate this. |
|
|
I and Bluestone went testing yesterday. I experienced too many random lags for it to be conclusive on my part. Bluestone reported no performance loss that was coincident with people joining/leaving. We were testing using this game: http://zero-k.info/Battles/Detail/265052 Note that using ZKL might be quickest way of procuring the game/map/engine, because while the game version is uploaded to rapid, it's not indexed. I've tried joining a TechA game that was going and they didn't report any fps loss as i was joining. |
|
|
"Bluestone reported no performance loss that was coincident with people joining/leaving." That wasn't exactly what I said - I only know as much as what I've said in the above post on this ticket. The mod version needed is available here: http://packages.springrts.com/builds/ but for some reason does not auto-dl from rapid. |
|
|
You had one spike for UnitAI but you didn't say anything about any consequential or previous spikes for join/leave being observed for whatever reasons. Thus you were not observing the thing, which is what i wrote. |
|
|
I wasn't aware of if/when players were joining/leaving (I don't even know find that out in ZK) and made no attempt to correlate to it. Your statement above says that I had attempted to correlate to it and had found no correlation, which is not true. E.g. It's entirely possible that a player joining/leaving/reconnecting/etc triggers a bug in "Tactical Unit AI". |
|
|
Ok i guess while it doesn't strictly say that (it's negation of observation) it can be read this way (observation of negation). Good that it got disambiguated. |
|
|
"And about that player leaves goes afk and back type of lag. This was definitely here with 91 I clearly remember that on my laptop. " -- ZK forums It seems that the issue was present previously, but is now simply exacerbated by something else. |
|
|
"20140601_164432_Bandit_Plains_v1_97.0.1-3-gda04dd2\ develop.sdf" when watching the replay, i didn't notice any lag from minute 7-9... |
|
|
not sure if this is a good way to measure: https://gist.github.com/abma/6ba967cb137c811449ed highest value i saw was 237 when replaying this demo, but it was for a single frame, nothing continously. |
|
|
hmmm, is speedcontrol maybe broken? with speedcontrol i mean the code that slows down the game when cpu load becomes to high. thats just a wild guess atm... |
|
|
watched a game big game with 23 players & ~10 specs, this is quiet common: [f=0010044] Connection attempt from <someuser> [f=0010044] -> Version: 97.0.1-3-gda04dd2 develop [f=0010044] -> Connection established (given id 31) [f=0010209] [Game::ClientReadNet] added new player Connetable with number 34 to team 0 [f=0010210] GameFrame() delay: 191 happened every (?) time a player joined. 191 imo is fine, also i read no complains about lags or sth. like that... not sure if 0,2sec lag is noticeable at all if its only a single gameframe. http://zero-k.info/Battles/Detail/266584 |
|
|
Dunno, it seems like you filter cases where a frame takes more than 0.1s to perform. FPS could easily drop from 90 to 10 without being detected here. |
|
|
I see large ping spikes (2 to 3 seconds) from many players when a spectator joins a large game. |
|
|
Also occurs with 97.0.1-18. |
|
|
Apparently heavy lag for many players when i joined http://zero-k.info/Battles/Detail/267186 at ~18:50. |
|
|
i didn't investigate, but i tested with a springie and local hosted games: /cheat /give 1000 armflea and then issued some commands to them. locally i get: [f=0000471] Warning: Waiting packet limit was reached for [AG]abma [packets dropped] on a springie hosted game i don't get this message?? hu? are Link* settings/limits on springies disabled or set to something insane high?! or don't i see this messages if i hit remotely such a limit? this could be useful to debug this maybe... creating a statistic with demotool / commands per frame maybe helps, too. with this limits disabled a custom widget could cause this for all players. |
|
|
From what i can see, the Springie instance running all of ZK atm has blank springsettigs.cfg. I assume that means it uses all-default settings. |
|
|
hm, zklobby seems to set LinkBandwidth = 0: https://code.google.com/p/zero-k/source/browse/trunk/other/ZeroKLobby/Resources/Conf/springsettings0.cfg not sure if spring-dedicated has some bw/packet/cmd-limit or how this affects cmd-limits. springlobby users shouldn't be affected by this as it doesn't disable/set any limit. |
|
|
According to this page that LinkBandwidth is not a setting. http://springrts.com/wiki/Springsettings.cfg |
|
|
indeed, LinkBandwith was removed some time ago, so this setting does nothing :-| @cleanrock: "...heavy lag for many players when..." lag or fps drop? this bug report mostly is about fps drop... |
|
|
@abma, gfrog was spec in game when i joined and said this: [GoogleFrog]cleanrock we're definately getting the load on player join here gfrog, can you describe what you "saw" when i joined ? |
|
|
When cleanrock started joining as a spectator I noticed a spike in CPU and ping for many players as shown on /info. |
|
|
Nothing about this report makes any sense. Literally *all* the server does besides very minor bookkeeping when a player joins to spectate or leaves a game is broadcast PlayerAdded or PlayerRemoved events to clients, which then inform Lua. NOTHING ELSE HAPPENS ON THE ENGINE / CLIENT SIDE, so before this goes any further and people possibly waste their time I want to see a ZK test with those specific callins (widgetHandler:Player* AND gadgetHandler:Player*) *disabled* or even better, extended with code that logs how much time each individual widget/gadget takes to execute them. |
|
|
and as predicted, only 1.5 years later: https://github.com/ZeroK-RTS/Zero-K/commit/3b77d543c53a5020dc49e3aaa014f0d55010760e |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2014-06-04 09:53 | Anarchid | New Issue | |
| 2014-06-04 11:29 | msafwan | Note Added: 0013195 | |
| 2014-06-04 11:29 | msafwan | Note Edited: 0013195 | |
| 2014-06-04 12:03 | silentwings | Note Added: 0013196 | |
| 2014-06-04 14:51 | abma | Note Added: 0013197 | |
| 2014-06-04 16:09 | abma | Status | new => feedback |
| 2014-06-04 16:09 | Anarchid | Note Added: 0013198 | |
| 2014-06-04 16:09 | Anarchid | Status | feedback => new |
| 2014-06-04 16:10 | Anarchid | Note Edited: 0013198 | |
| 2014-06-04 16:13 | abma | Note Added: 0013199 | |
| 2014-06-04 16:19 | Anarchid | Note Added: 0013200 | |
| 2014-06-04 16:42 | abma | Target Version | => 98.0 |
| 2014-06-05 20:09 | abma | Note Added: 0013206 | |
| 2014-06-05 20:10 | abma | Note Edited: 0013206 | |
| 2014-06-05 20:22 | silentwings | Note Added: 0013207 | |
| 2014-06-05 20:22 | silentwings | File Added: 20140601_164432_Bandit_Plains_v1_97.0.1-3-gda04dd2 develop.sdf | |
| 2014-06-05 20:23 | silentwings | Note Edited: 0013207 | |
| 2014-06-05 20:23 | Anarchid | Note Added: 0013208 | |
| 2014-06-05 20:25 | silentwings | Note Added: 0013209 | |
| 2014-06-05 20:26 | silentwings | Note Edited: 0013209 | |
| 2014-06-05 20:30 | Anarchid | Note Added: 0013210 | |
| 2014-06-05 20:51 | silentwings | Note Added: 0013211 | |
| 2014-06-05 20:52 | silentwings | Note Edited: 0013211 | |
| 2014-06-05 20:53 | silentwings | Note Edited: 0013211 | |
| 2014-06-05 20:59 | Anarchid | Note Added: 0013212 | |
| 2014-06-05 20:59 | silentwings | Note Edited: 0013211 | |
| 2014-06-06 10:13 | Anarchid | Note Added: 0013215 | |
| 2014-06-06 19:28 | abma | Note Added: 0013220 | |
| 2014-06-06 20:08 | abma | Note Added: 0013221 | |
| 2014-06-06 20:19 | abma | Note Added: 0013222 | |
| 2014-06-06 20:20 | abma | Note Edited: 0013222 | |
| 2014-06-06 21:56 | abma | Note Added: 0013224 | |
| 2014-06-06 21:56 | abma | Note Edited: 0013224 | |
| 2014-06-06 21:57 | abma | Note Edited: 0013224 | |
| 2014-06-06 22:04 | Anarchid | Note Added: 0013225 | |
| 2014-06-07 14:49 | Google_Frog | Note Added: 0013230 | |
| 2014-06-07 19:23 | Google_Frog | Note Added: 0013231 | |
| 2014-06-08 19:24 | cleanrock | Note Added: 0013233 | |
| 2014-06-08 20:46 | abma | Note Added: 0013234 | |
| 2014-06-08 20:47 | abma | Status | new => feedback |
| 2014-06-08 20:47 | abma | Note Edited: 0013234 | |
| 2014-06-08 20:47 | abma | Note Edited: 0013234 | |
| 2014-06-08 20:50 | abma | Note Edited: 0013234 | |
| 2014-06-08 20:52 | abma | Note Edited: 0013234 | |
| 2014-06-08 20:53 | abma | Note Edited: 0013234 | |
| 2014-06-08 20:54 | abma | Note Edited: 0013234 | |
| 2014-06-08 21:02 | Anarchid | Note Added: 0013235 | |
| 2014-06-08 21:02 | Anarchid | Status | feedback => new |
| 2014-06-08 21:07 | abma | Note Added: 0013236 | |
| 2014-06-08 21:08 | abma | Note Edited: 0013236 | |
| 2014-06-09 13:15 | Google_Frog | Note Added: 0013239 | |
| 2014-06-10 12:15 | abma | Note Added: 0013250 | |
| 2014-06-10 16:26 | cleanrock | Note Added: 0013257 | |
| 2014-06-11 09:57 | Google_Frog | Note Added: 0013265 | |
| 2014-06-11 15:49 | Kloot | Note Added: 0013268 | |
| 2014-06-11 15:50 | Kloot | Note Edited: 0013268 | |
| 2014-06-11 15:52 | Kloot | Note Edited: 0013268 | |
| 2014-06-11 15:54 | Kloot | Note Edited: 0013268 | |
| 2014-06-11 16:31 | Kloot | Note Edited: 0013268 | |
| 2014-06-11 16:31 | Kloot | Note Edited: 0013268 | |
| 2014-06-11 16:32 | Kloot | Status | new => feedback |
| 2014-06-15 00:52 | abma | Target Version | 98.0 => |
| 2017-01-07 12:32 | Kloot | Assigned To | => Kloot |
| 2017-01-07 12:32 | Kloot | Status | feedback => closed |
| 2017-01-07 12:32 | Kloot | Resolution | open => no change required |
| 2017-01-07 12:32 | Kloot | Note Added: 0017027 | |
| 2017-01-07 14:26 | Kloot | Note Edited: 0017027 |
