@omega_k2
I need evidence of that please in order to exact a change. Something measurable, something other than a personal account. Thanks
We know how our servers run and the general performance of them. How often there are dips ect and when those occur. We are in our own servers all the time enjoying the game and helping out players. When us server admins are saying there is a problem there is one. Even if we do not have a bunch of paperwork to back it up. Anyone that spends any regular amount of time monitoring their servers performance can tell when it has gone down or up.
Often these tick drops occur only for a short time and do not get measured by ns2servers. However as we are in our servers all the time or monitoring the tick from hlsw in real time we see when it starts to dip more and more often as builds go by. The computer is not getting tired and suddenly deciding to take a nap. No these issues get worse after new builds. Not all new builds but especially the resent new builds.
We can also say from observation that the performance has never been as good as it was a few months ago. Perhaps if we all kept a captains log it would be more apparent like:
"Star date 2013 12th month 18th day. Still seeing decreased performance in 261 with tick drops twice as far as seen in recent builds even on short games. The crew plans to make on offering to the NS2 gods today in hops that 262 will release today and fix the issues."
I don't need paperwork.
But i do need something more than just an inkling or someone saying "there a problem" .
Surely, if you are monitoring the tick rate that closely and that often, you could grab a screenshot. Bonus points if you can grab a server P_logall..
Ehm, look at the god damn server list. If you've been watching it before you'll see that plently of the server that did not have performance before 261 issues now have tickrate drops.
And if you've been watching your server usage reguarily, you'll also notice a significatent increase in CPU usage. The 18slt server now uses up to 90 (rarely 100%) cpu whereas it before used about 70% (rarely 80%) max.
And the 22 slot starts dropping tickrate now when capped (~130% cpu) to LOW numbers, like T15 in some situations. In 260 this didn't happen, and if there was an rare drop [rookie spam + descent or some other poor performance map] it was to ~T25 at most.
(note cpu % is cpu in relation to one cpu thread. 800% would be max)
Memory leaks. Look at the servers memory usage. It's not new, but it seems a bit worse then before. Starts at 700MB an goes slowly towards infinity, seen it end up at around 1600-1700mb, but usually by then either the server had it's daily restart or simply crashed.
You can see that the server hitches every few ticks and takes siginficantly longer to proccess a tick then usual. Sometimes those hitches are so large that they are noticable.
Now, those issues could have been found by UWE and the PTs, but it's quite obvious that they don't test some things properly at all (*whsipers* linux server release). And it's not something that affects only some servers, it seems all are affected.
I would suggest in the mean time the problem is any code that has changed since the build before reinforced lol. But do I have to run this P-logall on the server in advance? Can I grab this at any time? Is there a way to toggle it on and off?
The one thing that sux is that it is really hard to run commands on the server due to lack of rcon. Trying to use the web interface to do anything is not as good as it sounds and I am not even going to start that rant.
IronHorseDeveloper, QA Manager, Technical Support & contributorJoin Date: 2010-05-08Member: 71669Members, Super Administrators, Forum Admins, Forum Moderators, NS2 Developer, NS2 Playtester, Squad Five Blue, Subnautica Playtester, Subnautica PT Lead, Pistachionauts
edited December 2013
@Omega_k2 I am not denying there's an issue, i am saying we need info to go on.. and providing that green graph from one server (sample) with zero information about the hardware, connection, or mod installations, if any... is not useful or conclusive info.
Though, a larger sample size isn't as good or as useful as providing one singular p_log off a clean vanilla build of NS2 having troubles that are unacceptable. (sub 20 tickrate with hardware that was unchanged and previously capable of handling 30 consistently.) along with all the information about that server you can provide.
but looking at graphs.. comparing between those patches doesn't yield any notable or discernable difference.
Let me clarify this^
I see dips in tickrate before 261 and after. This Just confirms there's an issue but does not give any useful info to resolve the issue.
We need useful info from you server owners as i cannot personally reproduce this issue.. i don't have a server of my own and 24 players to test on (and therefore unable to grab a p_log)
sv_p_logall [minimal function time] Starts the capture of a profile log. for example p_logall 0 to capture everything
sv_p_endlog Ends the log started with p_logall and writes it to a file
Start it when you see things beginning to go south, to keep the file size down. ~10-15 minutes is acceptable (please compress the file before uploading it)
Try to capture the worst that you can.
Thanks
@Omega_k2 I am not denying there's an issue, i am saying we need info to go on.. and providing that green graph from one server (sample) with zero information about the hardware, connection, or mod installations, if any... is not useful or conclusive info.
Though, a larger sample size isn't as good or as useful as providing one singular p_log off a clean vanilla build of NS2 having troubles that are unacceptable. (sub 20 tickrate with hardware that was unchanged and previously capable of handling 30 consistently.) along with all the information about that server you can provide.
but looking at graphs.. comparing between those patches doesn't yield any notable or discernable difference.
Let me clarify this^
I see dips in tickrate before 261 and after. This Just confirms there's an issue but does not give any useful info to resolve the issue.
We need useful info from you server owners as i cannot personally reproduce this issue.. i don't have a server of my own and 24 players to test on (and therefore unable to grab a p_log)
sv_p_logall [minimal function time] Starts the capture of a profile log. for example p_logall 0 to capture everything
sv_p_endlog Ends the log started with p_logall and writes it to a file
Start it when you see things beginning to go south, to keep the file size down. ~10-15 minutes is acceptable (please compress the file before uploading it)
Try to capture the worst that you can.
Thanks
This problem surely has to be known at this point. The server hitching and tick rate drops have been going on for a very, very long time. Kind of silly to act like it's some new oddity that needs to be documented. Are the performance issues not the reason why most server operators overclock their hardware and we have all these silly server titles -- "4.6ghz overclock no lag!"?
Oh my god, this thread is a headache. So many people complaining but either doing nothing to help or touting the indefensible argument 'why should we help? Surely you know this and know how to fix it already!"
Why don't you try doing what ironhorse has asked you lazy bums!
@ns2isgood No, those silly server titles are because people want to attract players to their server, and are willing to mislead and/or misinform people in order to do so.
dePARAJoin Date: 2011-04-29Member: 96321Members, Squad Five Blue
From my point of view as a serveradmin i cant find a increase in cpu usage or memory leaks.
- One server.exe is still going up to 27-29% cpu usage like in all the builds before.
- One serverinstance use around 1 gig of ram wich is normal also.
I know what you really mean : you mean that 30 tickrate is low in general, which no one would disagree with - because you cant possibly be saying 1/30th or even 1/60th of a variance in server performance is "Unacceptable"... because that would just be oblivious to practical server performance for any game.
@omega_k2
Thanks for the plogs you sent in the PM, investigating now
This problem surely has to be known at this point.
OMFG.. read the first damn sentence of my last post *pulls hair out*
I did read your post and it has nothing to do with denying or admitting there is an issue. Would be kind of silly to try and deny such an obvious flaw in this game, the performance issues, which have gotten much better thankfully. I was making the implication that the problem has been known for a while, and asking for data and logs on the problem is where the comment "This problem surely has to be known at this point" was directed towards. I understand collecting data is how things get fixed quicker, but this problem has been ongoing since I started playing this game over 6 months ago.
NarfwakJoin Date: 2002-11-02Member: 5258Members, Super Administrators, Forum Admins, NS1 Playtester, Playtest Lead, Forum Moderators, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Supporter, Reinforced - Silver, Reinforced - Gold, Reinforced - Diamond, Reinforced - Shadow, Subnautica PT Lead, NS2 Community Developer
There's a lot more than one way the server tick rate can dip. So while I suppose it is accurate to say that "generic server stability problems have been a long term issue" that does nothing to address current problems. There have been a lot of server-side LUA error fixes checked in over the last six months including some very weird ones that were far worse than what has been reported in this thread.
What Ironhorse is getting at is that we are aware that something is going on, but we don't know exactly what the problem is. He's asking for the community's help providing hard data to single out the issue(s). There are a tremendous number of variables at play here and as not all servers exhibit the same problems or indeed any problems at all it's rather difficult for us to divine the nature of the issue on anecdote alone.
Sorry if that comes across as a bit snarky, but Iron's had a bit of a rough week and this thread was making me facepalm really hard.
Comments
I need evidence of that please in order to exact a change. Something measurable, something other than a personal account. Thanks
We know how our servers run and the general performance of them. How often there are dips ect and when those occur. We are in our own servers all the time enjoying the game and helping out players. When us server admins are saying there is a problem there is one. Even if we do not have a bunch of paperwork to back it up. Anyone that spends any regular amount of time monitoring their servers performance can tell when it has gone down or up.
Often these tick drops occur only for a short time and do not get measured by ns2servers. However as we are in our servers all the time or monitoring the tick from hlsw in real time we see when it starts to dip more and more often as builds go by. The computer is not getting tired and suddenly deciding to take a nap. No these issues get worse after new builds. Not all new builds but especially the resent new builds.
We can also say from observation that the performance has never been as good as it was a few months ago. Perhaps if we all kept a captains log it would be more apparent like:
"Star date 2013 12th month 18th day. Still seeing decreased performance in 261 with tick drops twice as far as seen in recent builds even on short games. The crew plans to make on offering to the NS2 gods today in hops that 262 will release today and fix the issues."
But i do need something more than just an inkling or someone saying "there a problem" .
Surely, if you are monitoring the tick rate that closely and that often, you could grab a screenshot. Bonus points if you can grab a server P_logall..
Ehm, look at the god damn server list. If you've been watching it before you'll see that plently of the server that did not have performance before 261 issues now have tickrate drops.
I made a topic on the issue, even though it got hijacked by who can't read apprently it does have some related replies:
http://forums.unknownworlds.com/discussion/133403/b261-server-performance#latest
And if you've been watching your server usage reguarily, you'll also notice a significatent increase in CPU usage. The 18slt server now uses up to 90 (rarely 100%) cpu whereas it before used about 70% (rarely 80%) max.
And the 22 slot starts dropping tickrate now when capped (~130% cpu) to LOW numbers, like T15 in some situations. In 260 this didn't happen, and if there was an rare drop [rookie spam + descent or some other poor performance map] it was to ~T25 at most.
(note cpu % is cpu in relation to one cpu thread. 800% would be max)
Memory leaks. Look at the servers memory usage. It's not new, but it seems a bit worse then before. Starts at 700MB an goes slowly towards infinity, seen it end up at around 1600-1700mb, but usually by then either the server had it's daily restart or simply crashed.
And if you want a nice recording, there is also this issue, although I didn't re-do the recording in build 261.
http://forums.unknownworlds.com/discussion/comment/2169848/#Comment_2169848
You can see that the server hitches every few ticks and takes siginficantly longer to proccess a tick then usual. Sometimes those hitches are so large that they are noticable.
Now, those issues could have been found by UWE and the PTs, but it's quite obvious that they don't test some things properly at all (*whsipers* linux server release). And it's not something that affects only some servers, it seems all are affected.
The one thing that sux is that it is really hard to run commands on the server due to lack of rcon. Trying to use the web interface to do anything is not as good as it sounds and I am not even going to start that rant.
I am not denying there's an issue, i am saying we need info to go on.. and providing that green graph from one server (sample) with zero information about the hardware, connection, or mod installations, if any... is not useful or conclusive info.
Though, a larger sample size isn't as good or as useful as providing one singular p_log off a clean vanilla build of NS2 having troubles that are unacceptable. (sub 20 tickrate with hardware that was unchanged and previously capable of handling 30 consistently.) along with all the information about that server you can provide.
Let me clarify this^
I see dips in tickrate before 261 and after. This Just confirms there's an issue but does not give any useful info to resolve the issue.
We need useful info from you server owners as i cannot personally reproduce this issue.. i don't have a server of my own and 24 players to test on (and therefore unable to grab a p_log)
@zeroibis
Start it when you see things beginning to go south, to keep the file size down. ~10-15 minutes is acceptable (please compress the file before uploading it)
Try to capture the worst that you can.
Thanks
I would say sub 29 is unacceptable...
This problem surely has to be known at this point. The server hitching and tick rate drops have been going on for a very, very long time. Kind of silly to act like it's some new oddity that needs to be documented. Are the performance issues not the reason why most server operators overclock their hardware and we have all these silly server titles -- "4.6ghz overclock no lag!"?
Why don't you try doing what ironhorse has asked you lazy bums!
@ns2isgood No, those silly server titles are because people want to attract players to their server, and are willing to mislead and/or misinform people in order to do so.
- One server.exe is still going up to 27-29% cpu usage like in all the builds before.
- One serverinstance use around 1 gig of ram wich is normal also.
No lua spam is decreasing the server.
Sorry, but i cant find any issues on my side.
Agreed, problems on a server in terms of lagging and rubberbanding arise from 29 and downwards in my experience.
@meatmachine
Thank you for seemingly being the only one comprehending my posts (am i not being clear enough or something?)
I know what you really mean : you mean that 30 tickrate is low in general, which no one would disagree with - because you cant possibly be saying 1/30th or even 1/60th of a variance in server performance is "Unacceptable"... because that would just be oblivious to practical server performance for any game.
@omega_k2
Thanks for the plogs you sent in the PM, investigating now
@omega_k2
That log was very useful, thank you.
Will get back to you soon with more
I did read your post and it has nothing to do with denying or admitting there is an issue. Would be kind of silly to try and deny such an obvious flaw in this game, the performance issues, which have gotten much better thankfully. I was making the implication that the problem has been known for a while, and asking for data and logs on the problem is where the comment "This problem surely has to be known at this point" was directed towards. I understand collecting data is how things get fixed quicker, but this problem has been ongoing since I started playing this game over 6 months ago.
What Ironhorse is getting at is that we are aware that something is going on, but we don't know exactly what the problem is. He's asking for the community's help providing hard data to single out the issue(s). There are a tremendous number of variables at play here and as not all servers exhibit the same problems or indeed any problems at all it's rather difficult for us to divine the nature of the issue on anecdote alone.
Sorry if that comes across as a bit snarky, but Iron's had a bit of a rough week and this thread was making me facepalm really hard.