Dancer Wrote:
> Henrik Nordstrom wrote:
> > Which means that if one cache_dir disk is gone, then Squid reduces the
> > cache to 0?
> >
> > /Henrik
>
> I think squid would simply fail to start if a cache_dir was
> inaccessible...which is far more likely (IMO) than it being accessible,
> but reading 0 bytes available.
I think Henrik's point is something like, when you get some types of SCSI
errors, that disk from that time on will still work for reads but will
return ENOSPC for all writes (even when the error was a simple timeout with
failed retry on some systems!). So every time squid decides to try and write
a file on the failed device, it reduces the cache size and decides to unlink
various files, and eventually removes all files including those on the
working drives.
David.
Received on Tue Jul 29 2003 - 13:15:51 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:51 MST