Hi,
On Fri, 10 Dec 1999, Kendall Lister wrote:
> On Thu, 9 Dec 1999, Nik Gibson wrote:
>
> > I have a production box running the squid proxy server. Squid keeps
> > dying with the following errors in the syslog.
> >
> > Dec 9 14:12:24 (squid): file_map_allocate: Exceeded filemap limit
> > Dec 9 14:12:24 squid[813]: Squid Parent: child process 814 exited with
> > status 1
>
> Your average_object_size is too large for your cache setup; check
> cachemgr.cgi's "Store Directory Stats" for details of the current state,
> and then check "General Runtime Information" to find out the actual
> average object size for your particular cache - if this is significantly
> different from the setting in squid.conf (probably 13 Kb), then consider
> changing it there.
>
Actually I think you'll find the average object is too small. Remember
that the filemap size is determined by
cache_size * 2 / ave_obj_size
So, the smaller the object size the more objects you can fit into the
cache and the more likely you are to run out of filemap entries. This also
probably indicates that L1 * L2 * L2 is >> the size of the filemap.
BTW the above formula allows for an ave_obj_size of down to 6.5kb before
overflow occurs. If the average size is less than that then the users
aren't looking at enough pictures :-)
Colin
Received on Thu Dec 09 1999 - 20:45:03 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:49:49 MST