[horde] Re: Using PHP Accelerators with Horde

[Daniel Eckl]

Am Montag, 14. Februar 2005 20:34 schrieb Andy Wright:
> > This problem is well known. Just search the list.
> >
> > mmcache stores binary data in shared memory and on disk. After restart /
> > reload it seems that the disk cache is corrupted somehow.
> >
> > You have two options.
> >
> > 1. Disable caching to disk. mmcache option shm_only. Bad, because then
> > scripts which are too big for SHM caching are not cached at all.
> >
> > 2. Keep the config and change the start/stop script to clear the mmcache
> > caching directory at every start / restart. Working for me since months
> > without problems.
>
> Thanks... I did come across one or two items on the list but none seemed
> to match the symptoms I'm seeing. The problem occurs without any
> restarts taking place so seems not to be the same as the problem you
> mention. It sometimes will works for many days and then fall over - no
> restarts, and plenty of memory left.
>
> I've allocated more than enough memory to ensure SHM never runs out so
> I'll try the shm_only option and see how that goes.
>
> Cheers,
> Andy.

I'm sure your apache _is_ restarted. One example is the log rotate script that 
sends a SIGHUP to the apache process to force it to reopen the log file. I 
have edited my logrotate script to empty the mmcache dir, too.

But of course you have the shm_option to test if that fits your needs.

Best,
Daniel