Server Dedicated Server Crush bug

[rallfo]

We run Linux across all our machines too...

I was just about to stress test it on our main with 32 players versus 32 patriarchs as well...

 

[[TW]Ramm-Jaeger]

Is the fix only for windows servers?

Linux is affected too.

This fix is only for windows servers, everyone was saying there was nothing wrong with the Linux servers. If they are having the issue as well, we can put the fix out for them too.

 

[mastertheknife]

CentOS linux here, affected aswell.
So yeah, a fix would be welcome

 

[Z-MAN]

-beta kfwinserverfix1006

thx.
The change in kfchar.u and engine.u was confirmed.
It executes by 7 threads and it tests.

 

[rallfo]

everyone was saying there was nothing wrong with the Linux servers. If they are having the issue as well, we can put the fix out for them too.

Yes please!

 

[arramus]

We now have a fix for this crash up on HLDS. It is currently behind a beta password until we can verify that it is ready to be released fully. If you would like to test out this fix, please add the following to your HLSD commandline to grab the update:

-beta kfwinserverfix1006

Please post on here if you are testing the fix and let us know how it works for you.

I'm testing the fix and will report back the results over the next few days.

 

[Z-MAN]

ummmm....

Is this a beta file?
kfchar.u ： 300,920 byte

Isn't it this bug?
>Adjusted the spawning of the Patriarch boss character to prevent cases where the boss was spawning too close to players.


Log: PlayAnim: Sequence 'Jump' not found for mesh 'Patriarch_Freak'
ScriptLog: CHOOSEATTACKSERIAL in state InitialHunting enemy NONE old enemy NONE CALLING BYTE 26
:
LOOP
:
ScriptLog: CHOOSEATTACKSERIAL in state InitialHunting enemy NONE old enemy NONE CALLING BYTE 26
Critical: BossZombieController KF-ElementarySchool.BossZombieController (Function Old2k4.MonsterController.ChooseAttackMode:003E) Runaway loop detected (over 10000000 iterations)
Exit: Executing UObject::StaticShutdownAfterError

 

[Plattfuss]

Same here (linux)!

Posted this days ago, so yes, we need the fix also for linux!

 

[kingkong]

so how goes the fix, anyone? I didn't want to install this at first, since I was yet to come across any instance of this bug. I just saw it for the first time today (xp-x86), in a case where I was in a spot that the pat probably did not have a path to. I was "testing" a map on my server (KF-The_Rock_beta1), sitting on a ladder in an elevator shaft. as those of you who have played this map probably know, the zeds here have trouble finding players in this area.

so z-man I think what you're looking at could be their previously stable pat-spawning routine, although not always in ideal positions was probably less prone to crashing on pathing errors, so it looks like it has been reverted(?) along with whatever solution they came up with to work around this, just a guess.

and this error -
Log: PlayAnim: Sequence 'Jump' not found for mesh 'Patriarch_Freak'

is probably the most common one I see in my ucc output, pretty much every round that is played spams it constantly all throughout the game.

along with -
Log: PlayAnim: Sequence 'Hit[R|L|B]_Katana' not found for mesh 'british_riot_police_I'

and -
Log: PlayAnim: Sequence 'DeathB' not found for mesh 'british_riot_police_I'

every time this character is in use. never causes any crashes though, but no doubt the game would probably run a lot smoother without these constant errors.

 

[arramus]

I've had 2 servers up using the beta kfwinserverfix1006 fix for a couple of days now without any crash incidents. The servers have been well populated over that time and I've made a conscious effort to join them periodically as well.

Overall gameplay in these servers doesn't feel quite as smooth as prior to the latest updates with noticable player warping on a few occasions. Pre caching of maps is also a lot longer when joining the servers as well.

 

[Starfi]

This fix is only for windows servers, everyone was saying there was nothing wrong with the Linux servers. If they are having the issue as well, we can put the fix out for them too.

Please patch the Linux Servers ...

My KF Server (Linux) crashes eyery day !!

 

[Z-MAN]

Is this a beta file?
kfchar.u ： 300,920 byte

Error: KFGameType KF-AquaStudy-V2Final.KFGameType (Function KFMod.KFGameType.MatchInProgress.SetupPickups:01D7) Accessed array 'WeaponPickups' out of bounds (0/0)
Warning: KFGameType KF-AquaStudy-V2Final.KFGameType (Function KFMod.KFGameType.MatchInProgress.SetupPickups:01D7) Accessed None 'WeaponPickups'
NetComeGo: Close TcpipConnection *.*.*.*:* 09/01/09 00:24:37
ScriptLog: CHOOSEATTACKSERIAL in state InitialHunting enemy NONE old enemy NONE CALLING BYTE 26
:
ScriptLog: CHOOSEATTACKSERIAL in state InitialHunting enemy NONE old enemy NONE CALLING BYTE 26
Critical: BossZombieController KF-AquaStudy-V2Final.BossZombieController (Function KFMod.KFMonsterController.ExecuteWhatToDoNext:0092) Runaway loop detected (over 10000000 iterations)
Exit: Executing UObject::StaticShutdownAfterError

 

[Z-MAN]

The phenomenon was completely reproduced.

The server crashes when failing in the player's search.
Patriarch spawn or Patriarch run.

Example:
It fails in the search in waiting on the stand of the KF-Concert_Hall.
The server crashes when Patriarch spawn waiting on the stand.

solo crush log.

Build KillingFloor_Build_[2009-02-09_10.82]

OS: Windows XP 5.1 (Build: 2600)
CPU: GenuineIntel PentiumPro-class processor @ 3171 MHz with 2046MB RAM
Video: ATI Radeon HD 4800 Series (6983)

BossZombieController KF-Concert_Hall.BossZombieController (Function Old2k4.MonsterController.EnemyVisible:0001) Runaway loop detected (over 10000000 iterations)

History: FFrame::Serialize <- AActor:rocessState <- Object BossZombieController KF-Concert_Hall.BossZombieController, Old State State KFChar.BossZombieController.InitialHunting, New State State KFChar.BossZombieController.InitialHunting <- AController::Tick <- TickAllActors <- ULevel::Tick <- (NetMode=0) <- TickLevel <- UGameEngine::Tick <- Level KF-Concert_Hall_Final_Beta <- UpdateWorld <- MainLoop <- FMallocWindows::Free <- FMallocWindows::Realloc <- 4C576C61 0 FArray <- FArray::Realloc <- 0*2 <- FMallocWindows::Free


fix it at once.

thank you for a test by a Japanese community.

 

[Plattfuss]

Please fix that bug for linux, its very boring to start the server every 1-2 hours!

 

[arramus]

A good 3 days without a crash now and I've played multiple end waves with the Patriarch albeit in custom maps. I haven't attempted to test Z-MAN's hypothesis about the server crashing when the players are elevated above ground? when the Patriarch is searching for them but will give it a go.

 

[Z-MAN]

A good 3 days without a crash now and I've played multiple end waves with the Patriarch albeit in custom maps. I haven't attempted to test Z-MAN's hypothesis about the server crashing when the players are elevated above ground? when the Patriarch is searching for them but will give it a go.

The crash happens easily in a lot of places.
Roomette in kf-farm.
Corner in kf-offices.

It crashes when staying 1 player.

player 1 ok
player 2 ok
player 3 error -> crush.

 

[rallfo]

Testing things like this is all about edge cases, players floating above the ground, one hundred Patriarchs, reverse gravity...

We can do that on our server if the beta fix is rolled out for Linux!

Also I trust the final fix _will_ make it to the Linux platform at any rate...?

 

[kingkong]

infinite loop bug still exists, even with beta patch. the server is running on xp-x86, intel p4 northwood 2.26 ghz / 845e, 1gb ddr.

CHOOSEATTACKSERIAL in state InitialHunting enemy NONE old enemy NONE CALLING BYTE 26

with runaway loop just like z-man has posted above.

I just had a chance to test the exact same case I found it in before, and it crashed in exactly the same way, before he could spawn and do the 3rd person view.

to be clear, my method of reproducing this bug is not actually in an unreachable spot, just one that the ai is not easily able to negotiate.

 

[arramus]

My public mixed map server's gone four days without a crash and it's been well populated with players using custom and retail maps. I purposely haven't restarted this one for the duration.

I didn't get a chance to reproduce the loop bug scenarios and will give that a run through later today.

 

[Z-MAN]

If it is not possible to fix, it is necessary to rewind it to an old algorithm.

A new algorithm appears near the player and is bad.
An old algorithm is better.