>
> Hi all,
>
> recently i had huge performance problems on one of my squid
> boxes (2.5stable4, Solaris8 non-chroot). It truned out that
> almost every object request produced an entry similar to the
> following in the chache.log:
>
> 2004/07/13 07:40:34| ctx: enter level 0: 'http://www.google/'
> 2004/07/13 07:40:34| unlinkdUnlink: write FD 20 failed: (32)
> Broken pipe
> 2004/07/13 07:40:40| ctx: exit level 0
>
> After i did a restart of the squid, the problem seems to be
> gone for now. Still, this solution is kind of unsatisfying,
> so i searched the archive for the above error. Apparently
> the unlinkd seems to have crashed, but the archived posts
> lack an explaination what may have caused the crash. Is there
> a simple way to determine why the unlinkd failed?
>
Perhaps the process left a core dump somewhere.
If so , using debuggers e.d. stack trace info could be gathered.
Also check squid's cache.log, in those circumstances, perhaps
more info can be seen over there.
Check system logs too, watch for any errors concerning process handling.
And watchout for more general error messages related to disk/io e.d,
(for instance).
M.
Received on Tue Jul 13 2004 - 02:12:58 MDT
This archive was generated by hypermail pre-2.1.9 : Sun Aug 01 2004 - 12:00:01 MDT