[Resolved] Warning: system time went backward by 1 seconds

[NIGathan]

I used the hldsupdatetool.bin to setup a clean unmodded tf2 server on a CentOS 5 32bit machine with the following specs:

Quote:Processor #1 Vendor: AuthenticAMD
Processor #1 Name: AMD Athlon™ Dual Core Processor 4850e
Processor #1 speed: 2500.301 MHz
Processor #1 cache size: 512 KB
4GB RAM

When I run the srcds_run it starts fine and I can connect, but my ping will rapidly change from anywhere from 100 to 2000. And the lag is unbearable. While its running, it constantly outputs "Warning: system time went backward by 1 seconds" about once every second.

Quote:nigathan@this.is.where.all.the.files.are.hosted.gtfo.justca.me [~/srcds_l/orangebox]# ./srcds_run -console -game tf +map ctf_2fort +maxplayers 12 -autoupdate
Auto detecting CPU
Using AMD Optimised binary.
Server will auto-restart if there is a crash.
Could not locate steam binary:./steam, ignoring.

Console initialized.
Game.dll loaded for "Team Fortress"
Particles: Missing 'particles/error.pcf'
maxplayers set to 12
Unknown command "r_decal_cullsize"
maxplayers set to 12
Unknown command "startupmenu"
Network: IP 212.117.169.222, mode MP, dedicated Yes, ports 27015 SV / 27005 CL
ConVarRef room_type doesn't point to an existing ConVar
Executing dedicated server config file
[S_API FAIL] SteamAPI_Init() failed; unable to update local steamclient. Continuing with current version anyway.
--------------------------------------------------------
sv_pure set to 1.
Note: Changes to sv_pure take effect when the next map is loaded.
--------------------------------------------------------
exec: couldn't exec banned_user.cfg
exec: couldn't exec banned_ip.cfg
exec: couldn't exec ctf_2fort.cfg
Adding master server 68.142.72.250:27011
Adding master server 72.165.61.136:27011
Connection to Steam servers successful.
VAC secure mode is activated.
Warning: system time went backward by 1 seconds
Warning: system time went backward by 1 seconds
Warning: system time went backward by 1 seconds
Warning: system time went backward by 1 seconds
Warning: system time went backward by 1 seconds

There is also an srcds setup with css and it doesnt have this problem. There is also an ircd running from this same account and it links with another ircd and never has any time syncing errors. So its most likely not a server issue, however I could be wrong.

Any suggestions would be much appreciated.

Thanks

[BehaartesEtwas]

have a look at the system log (less /var/log/messages) and the kernel messages (dmesg), if there are any abnormal messages there. also you might want to check, if the system clock really doesn't run smooth, e.g. try running the following command:

while true; do date; sleep 0.1; done

this will run "date" even 0.1 seconds, so you can see, if the system clock does any strange things.

do you have something like ntp running?

[NIGathan]

I see ntp running in htop, yet have no idea what it is.. Another user on this server probably uses it.

When I run that command it doesnt seem to be correct.. The time never actually goes backwards, however it doesnt increase by one second every time.

Quote:nigathan@this.is.where.all.the.files.are.hosted.gtfo.justca.me [~]# while true; do date; sleep 0.1; done
Wed Oct 14 09:26:09 CEST 2009
Wed Oct 14 09:26:09 CEST 2009
Wed Oct 14 09:26:10 CEST 2009
Wed Oct 14 09:26:12 CEST 2009
Wed Oct 14 09:26:12 CEST 2009
Wed Oct 14 09:26:12 CEST 2009
Wed Oct 14 09:26:12 CEST 2009
Wed Oct 14 09:26:13 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:14 CEST 2009
Wed Oct 14 09:26:15 CEST 2009
Wed Oct 14 09:26:15 CEST 2009

Edit: Nevermind.. Just re-read your comment and saw you mentioned it would do it every 0.1 seconds

Also, I looked for those log files, and couldnt find them, can you be a bit more specific on their locations?

Well I went ahead and killed the ntp process, and it appeared to slow the time between each warning, but now it goes back 2 seconds much more often. When I joined it seemed to be the slightest bit faster, and my ping rarely exceeded 1000, still unplayable tho..

[BehaartesEtwas]

but the clock doesn't run smooth anyways... it should print ~10 times the same second!

/var/log/messages -> this is the main log file of linux
dmesg -> this is a command to print all kernel messages

both probably only accessible as root...

did you change anything in the kernel? did you try different clock sources (see http://wiki.fragaholics.de/index.php/EN:Linux_Kernel_Optimization#Playing_around )?

also I have seen PCs with very bad clocks, but not that bad. So I would guess it's a software problem, but it's not impossible that some crystal is broken on your motherboard...

[css]

Is it virtual server?

Nevertheless, try stopping the NTP daemon. NTP is clock synchronizing system that keeps the clock in correct time. Use this command:

/etc/init.d/ntp stop

You can also try:

service ntp stop

[NIGathan]

Nope, not a virtual server. Just a jailed ssh account. I tried killing ntp from htop earlier, and that didnt help much at all.

BehaartesEtwas: We just recently installed a new kernel, and we had to recompile everything, but nothing else seems to be having any issues, and the time is always correct on our BNC, so I really do not think its a physical time error, if that were the case the srcds probably wouldnt even realize that it was running slower. But Im putting my money on it being a kernel issue, cause I had to work with my custom unrealircd for a while just to get it to recompile properly afterward..

Would it be possible to install the tf game files onto the previously installed css srcds, and just run one or the other? And if so, could I merely mv the files over to the other installation dir, so I dont have to go through the entire process of downloading it again?

Also thanks for that link, this isnt my server so Id rather not break anything so Im gonna wait till someone with a bit more knowledge of the server wakes up and Ill let him try it.

[BehaartesEtwas]

what is a "jailed ssh account" exactly? can you try running the server using a normal linux account without any fancy stuff?

I would also bet that this is a kernel problem. You might want to double check the configuration. Usual operations on a linux system might not be affected by clock problems, and if ntp synchronizes your clock permanently you might not even notice any problem.

btw: the clocksources mentioned in my howto are not dangerous to change. worst case scenario is that your server crashes (I never have seen this and I tried all options), but the original clocksource will be restored after a reboot in any case.

[NIGathan]

A jailed ssh account is a normal ssh account that only has access to a specific directory. I also have root access, but I like to keep certain things running from different accounts so I can easily sort by user in htop to find processes.

And alirhgt, Ill try changing the clocksource now

Edit:
Well changing the clocksource from tsc to hpet fixed the time errors, but now when I try to connect to the server I get disconnected with this message: "Disconnected: Server uses different class tables."

Edit:
I tried all options for the clocksource, and all of them seem to get rid of the original problem, but they all also cause the different class tables problem..

Edit:
After some googling, I found that I just needed to update the server, so I did that and everything works great! No lag whatsoever now! Thanks a lot BehaartesEtwas +rep for you