"No UDP Reply"

[Critter]

Hi All,

I am running TeamSpeak Server 1.2.2 for Linux.

Periodically, the program seems to crash, especially if there are a high number of people connected (12 or more), but sometimes it seems to go down when no one is connected.

If people are connected, they are disconnected and given an error to the tune of "No UDP Reply from server". When people attempt to connect after this happens, they receive the same error and are unable to connect.

When I have investigated this, I found nothing in the server log, and nothing in the system log indicating a failure. "ps -aux" reveals that there are still the normal 6 processes running as they should.

This happens on another 1.2.2 Linux server that I frequent, and it's possible that it's a fairly common error. If there were any FAQs or threads pertaining to this, they seem to have expired.

I'd appreciate the advice of some of you experienced administrators who have successfully diagnosed and dealt with this problem, or from you developers who have pity on poor little admins like myself :).

Critter

[N. Werensteijn]

It sounds like the server is having problems sending udp packets to you.

This can be because of
1: server's upload, or your download is saturated
2: ping is greater than 3000 ms
3: Other reason

It's a balance we had to make between a more stable connection or faster disconnection detection.

[Critter]

I considered this, but I'm on the same local network the server is on. My ping to it is less than 10ms, and I find it hard to believe that 650Bps x 12 clients would saturate a 100Mbps ethernet network :).

I did notice that the servers swap space and memory were chock full (according to "vmstat" and "free"). I had set up cron to restart the server every day, thinking I could free my RAM, however this didn't work.

I have now set up the server to reboot everyday, thinking this will flush the swap space and free up all my memory, but I still don't know if this is really addressing the issue. I'll let you know if it seems to work or if it doesn't.

Is there anything else that I might try? Rebooting the server is a long and painful task, and I'd rather not do it if I don't need to :). Thanks!

Edit: I might also add that everyone gets disconnected at the same time; the server simply dies, but you'd never know it (logs, ps, etc. seem fine) until you tried to connect.

I realize this product is at it's end of life, and I'm trying to push v2.0, but it doesn't seem as lean as v1.2 and when the server fills up I have BW issues.

[mintyfresh]

I'm experiencing the same problem and thought it was only because I was running TSS 1.2.2 as a Linux compatible application under OpenBSD 3.1.

Critter, please tell me what operating system you are using for TSS. I was thinking that switching to a native Linux OS would resolve the problem.

The issue for me is similar. The server is up and running (on a T-1 circuit) and then *BAM* no one else can get in. The people still connected are ok, but no new client connections can be established.

If you leave the server when it's in this state, you can't get back in. I agree with Critter in that it doesn't appear to be a scalability problem.

In fact, when it gets into this state, TSS won't terminate in an orderly fashion. SIGTERM is ignored and doesn't cause the server to shutdown. SIGKILL (-9) is the only way to stop it when it gets into this state.

Nothing in the error logs at all.

I'm willing to offer my help to debug and/or fix this problem. I'm a professional software developer with over 14 years of experience in building networking application on Unix platforms.

I know that among the gaming community I play with TS 1.5 is still perferred over TS 2.0. So, I think it still has a lot of life.

[Critter]

After doing a more intensive search of the forums I discovered a post from Niels stating that this is a bug, and will not be repaired. I am slowly forcing my users to migrate to TeamSpeak 2.0... it gets better after every developer patch and is rock solid as far as stability is concerned.

In the meantime the daily reboots are very effective at minimizing the effects of the TeamSpeak 1.2.2 bug.

I am running Mandrake Linux 8.2 with a 2.4.19 kernel. I'll be upgrading to Mandrake 9.0 once I have an opportunity to take the server down for a day.

[ScratchMonkey]

If you use iptables with connection tracking, note that the tracking system will "forget" any UDP "connection" that hasn't seen activity in 3 minutes. What this means is that you must allow all traffic into and out of your UDP port separately. (The more common practice is to allow traffic in the initiation direction (ie. to the server) and use connection tracking to automatically allow the replies, but this doesn't work with TSS1 because it leaves the connection idle too long.)

[rog2054]

A question which i'm sure has been asked a 1000 times but i cannot find the answer...

Can windows 1.5 clients connect to a linux TS2 server?

The reason is i need to use 1.5 clients due to compatibility issues with some online games.
But also need to use the linux v2 server because of the v1.22 bug as mentioned at the start of this thread.

(incidently, does this bug affect just v1.22 or all v1.x?)

[ScratchMonkey]

Alas, no, the 1.x and 2.x protocols are quite different, so you can't mix and match clients and servers.