Re: [squid-users] WARNING: failed to unpack meta data

The first is a symptom of your OS kernel running out of memory.

The second is a symptom of a corrupted cache, most likely caused by
Squid crashing due to the kernel memory problems..

If the second problem does not go away automatically you could try to
manually purge the corrupted objects from the cache. First find which
URLs it is complaining on (search for SWAPFAIL in access.log), then
PURGE these URL's from the cache.

Regards
Henrik Nordström
Squid Hacker

Mike Kiernan wrote:
>
> Had some strange problems with one of our lvm caches
> over the weekend (user's uploaded pages failing to refresh).
>
> The cache crashed on Friday night
> Sep 21 20:33:09 squid[1499]: comm_poll: poll failure: (12) Cannot
> allocate memory
> Sep 21 20:33:09 squid[1499]: Select loop Error. Retry 1
> Sep 21 20:33:09 squid[1499]: comm_poll: poll failure: (12) Cannot
> allocate memory
> Sep 21 20:33:09 squid[1499]: Select loop Error. Retry 2
>
> Sep 21 20:33:13 squid[980]: Rebuilding storage in
> /var/spool/squid-f(DIRTY)
> Sep 21 20:33:13 squid[980]: Rebuilding storage in /var/spool/squid-f-1
> (DIRTY)
>
> a little later lots of these:
>
> Sep 21 20:38:15 kernel: eth0: card reports no resources.
>
> The cache was reportedly returning passed-sellby pages on Saturday,
> but appears to have righted itself by Sunday [very vague I know,
> but nothing in the logs, just vague complaints to go on].
>
> Since then up until now we're getting the following message every few
> minutes:
>
> Sep 21 21:24:18 squid[980]: WARNING: failed to unpack meta data
>
> Message comes from storeClientReadHeader() callback.
>
> What exactly does the message mean, and does it have any implications?
>
> Thanks.
>
> --
> Michael Kiernan
> Onet.pl S.A.
> Krakow, Poland
> http://www.onet.pl/
Received on Tue Sep 25 2001 - 00:21:40 MDT