[tor-relays] Ubuntu is killing tor when getting low memory

[David Stainton]

Hi Geri!

You may adjust the Linux OOM killer's settings on a per process basis
with the proc fs; see here:
http://askubuntu.com/questions/60672/how-do-i-use-oom-score-adj

If you have multiple numa cores then it also might be helpful to set
the process to use numa interleaved memory
instead of just it's local numa memory bank... see numactl for more
information about that. Jeremy Cole wrote
an interesting article about tuning mysql with numactl a while back.

Some people advise keeping a sacrificial lamb process that gets
oom-killed first.

Also I always like to disable swap completely (e.g. swapoff -a)... but
sometimes I meet sys admins that
try to argue that having some swap is a good idea; i am not convinced.
swap is so 1992.

I didn't explain everything or go into all the details here... so feel
free to ask me questions if you have some
problems or if something is unclear.

Cheers!

David


On Sat, Feb 1, 2014 at 12:41 PM, toxi roxi <toxiroxi at gmail.com> wrote:
> Hi Folks,
>
> im running many relays which are doing a nice job. But i have also some
> smaller relays outside which have just 256mb.
> They are performing quite quell - but on one relay im facing issues which i
> think you may help me.
>
> this relay seems to ran out of memory from time to time (2 to 5 days) and i
> found out via dmesg, that ubuntu itself is killing the process to free up
> memory.
> the swap is used, but far away from full.
>
> as this stops tor immediately its bit annoying that i need to restart the
> releay without any reason.
>
> is there any way to tell ubuntu or to prevent that the tor process is killed
> for freeing up memory?
>
> i dont care about other processes as this machine is for tor purposes only.
>
> thanks for your expertise!
> Geri
>
> _______________________________________________
> tor-relays mailing list
> tor-relays at lists.torproject.org
> https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
>