I have a very strange and mindboggling issue. I have no clue what’s so ever what could be the cause or how to fix it. So, I’m open for any ideas, even if they seem farfetched at this point. I’m not an experienced Linux user, so please keep that in mind when trying to explain or giving instructions on how to do something. If you need more information to help me with this problem, please tell me how to get it and I will try.
Linux version 5.8.3-xanmod1 ([email protected]
) (gcc-10 (Debian 10.2.0-5) 10.2.0, GNU ld (GNU Binutils for Debian) 2.35) #0~git20200822.3996034 SMP PREEMPT Sat Aug 22 03
5:03 UTC 2020
Intel Xeon E-2288G
500GB NVMe SSD
So, we bought some new hardware and we are hosting CSGO game servers on them. On this new setup we went from Gentoo to Debian 10 with Xanmod. The servers are still on the same network as before and we have fresh installs of everything. Before the move, we had no issues. We are running 7 CSGO servers each with its own IP and all using the port 27015.
This is a bit strange and hard to explain. But every night, after 6PM ish, CSGO servers lose connection in some way. It looks very random. It happens across 7 machines with same setup. They don’t loose connection at the same time. One CSGO server might go down at 7.35 and next one at 8.52 and so on.
When it happens, its like the CSGO server in question goes into a LAN/OFFLINE mode or something. I can send rcon commands and receive console data via HLSW (game tool on my computer). But at the same time HLSW tries to ping the server and it gets time out. Players cannot join the server at all once it enters this “state” and players on the server disconnects.
There is a command in CSGO called “net_start” (this command will start multiplayer network sockets) and when I use it, the server goes online again and the problem goes away.
We had this problem ever since we installed the new machines, going 6 nights in a row where 4-7 CSGO servers randomly goes offline one by one and sometimes multiple servers within minutes. It always happens after 6PM (CSGO servers auto restarts at 5am). They run all day filled with players without issues. I did notice that high slot servers (16-18 players) goes down way more often than low slot servers (10 slots). In fact, we had about a total of 30 CSGO servers that went down and only one of them was a 10 slot. But once the servers go down, they could be empty so the number of current players does not seem to be a factor. Also had reports that players notice “lag” right before or during the disconnects.
What we manually installed:
Tuned - https://tuned-project.org/
We also run crontab without, not running anything that would explain the issue. We don’t use iptables. And our hoster says there are no issues with switch/routers. The old machines are taken down. And he flushed ARP tables.
I can’t find ANY errors in any logs. Not in CSGO and not for Debian/kernel/system. Except for these:
[71643.054913] TCP: request_sock_TCP: Possible SYN flooding on port 27015. Sending cookies. Check SNMP counters.
[60876.945710] perf: interrupt took too long (2501 > 2500), lowering kernel.perf_event_max_sample_rate to 79500