[rabbitmq-discuss] Silent crashes, and "killed by ABRT signal"

Emile Joubert emile at rabbitmq.com
Wed Dec 4 10:22:19 GMT 2013


It looks like the out-of-memory killer killed the Erlang VM because the
OS was under memory pressure.

On 04/12/13 09:53, carlhoerberg wrote:

> Dec  4 09:47:39 ip-10-34-235-83 kernel: [7494791.746025] beam invoked
> oom-killer: gfp_mask=0x201da, order=0, oom_adj=0, oom_score_adj=0

> Dec  4 09:47:40 ip-10-34-235-83 kernel: [7494791.746225] Free swap  = 0kB
> Dec  4 09:47:40 ip-10-34-235-83 kernel: [7494791.746227] Total swap = 0kB

Not having any swapspace won't help.

> Dec  4 09:47:40 ip-10-34-235-83 kernel: [7494792.224468] Out of memory: Kill
> process 8403 (beam) score 870 or sacrifice child
> Dec  4 09:47:40 ip-10-34-235-83 kernel: [7494792.224485] Killed process 8403
> (beam) total-vm:1228328kB, anon-rss:524748kB, file-rss:0kB
> Dec  4 09:49:22 ip-10-34-235-83 kernel: [7494894.351834] TCP: Possible SYN
> flooding on port 5672. Sending cookies.  Check SNMP counters.

It is possible to configure how the kernel deals with low memory in
overcommited situations. You should also configure some swapspace. If
you have any other services on the same OS that use alot of RAM then try
moving them to a different server.



-Emile


More information about the rabbitmq-discuss mailing list