[Savannah-hackers] Re: [savannah-root] load increase on subversions.gnu.org

[Mathieu Roy]

Vincent Caron <address@hidden> a tapoté :

> Mathieu Roy wrote:
>> The problem we encounter may well be due to the broken hard disk we
>> still have on the server.
>> When rearranging the partitions because we were severely running out,
>> I had to reactive hda for tmpcvs.
>
> That could also explain those processes obviously spending too long
> time in I/O, indeed.
>
> BTW, I swear I saw free reporting an oscillating 0-5MB free on swap
> space 2 hours ago, while load was averaging 25 (yeees :)). Could be a
> consequence of the mentioned disk pb: too many slow CVS processes
> stacking up and consuming a lot of memory.

It could. It was this kind of troubles that forced Loic to almost shut
down hda.

But I've no knowledge in black magic and we need these partitions I
added on hda.

>> And the problem we have are indeed with tmpcvs. There was no problem
>> at first, but since it now occurs every days, it may well be an
>> hardware problem.
>
> Was not this filesystem mounted as ramfs at the beginning ?

I can't remember, you have to check the /root/ChangeLog

>
>
>> The current solution is see would be to have a cronjob that check for
>> process that take up to one certain amount of RAM, kill the
>> incriminated process, run sv_nurse_tmpcvs
>
> We can try to limit the allocatable memory for a CVS process. However
> I only know about bash's ulimit which is not suitable for cvspserver
> usage.

I have no idea about that either.


-- 
Mathieu Roy

  +---------------------------------------------------------------------+
  | General Homepage:           http://yeupou.coleumes.org/             |
  | Computing Homepage:         http://alberich.coleumes.org/           |
  | Not a native english speaker:                                       |
  |     http://stock.coleumes.org/doc.php?i=/misc-files/flawed-english  |
  +---------------------------------------------------------------------+