[Dev] Server crashes can be OOMs
Esteban Carnevale
alfplayer at mailoo.org
Fri Mar 18 14:19:29 GMT 2016
On 2016-03-17 15:29, fauno wrote:
> Esteban Carnevale <alfplayer at mailoo.org> writes:
>
>> On 2016-03-17 14:00, Esteban Carnevale wrote:
>>> Memory on the server is normally ~ 50 %. Sometimes Labs Ruby
>>> processes
>>> are executed which use ~ 30 % total. I've seen Parabolaweb use ~ 10
>>> %,
>>> so if these two, and possibly others, take memory at the same time,
>>> the
>>> server can go out of memory. Is this acceptable for Labs?
>>>
>>> Examples of "sar -r" and "top -o %MEM" with high Labs memory usage on
>>> the file attached.
>>
>> Parabolaweb alone taking 50 % later today (shown on attached file).
>> It's an OOM waiting to happen. I think I'll reboot it any minute now
>> as I lack a better idea.
>
> btw as a quick fix you can configure monit to monitor parabolaweb's by
> pid and restart it when ram usage is over certain threshold.
It reached 90 %. I restarted parabolaweb some minutes ago. Let's see
how much it helps.
--
Esteban Carnevale <alfplayer at mailoo.org>
More information about the Dev
mailing list