RE: effect of lower store_avg_obj_size setting

I am runnning 2.2stable4 on top of a linux 2.0.36 kernel.

You are right, my average object size was only around 6 K. I changed
it to 8 in my config.

The only thing that I don't understand is that this server
is not very busy. It only has about 8 light simultaneous users, and
I would think that the 486 with 32MB could handle that load
with no problems. The server doesn't provide any other services,
except for packet filtering.

One of the other reasons why I am asking this question is I want to
give my boss a little bit of reassurance that I have fixed the problem
with the config change.

thanks Henrik,

geoff

>
> Geoff Nordli wrote:
> >
> > Why does reducing the size of the lower store_avg_obj_size setting
> > get rid of the error "FATAL: file_map_allocate: Exceeded
> filemap limit".
>
> Upon startup Squid allocates a static bitmap for keeping
> track of which
> cache_dir files are being in use. The size of this bitmap is
> determined
> by the cache_dir size and store_avg_obj_size.
>
> > I would think that if you have enough cache (100MB) you still have
> > 7700 objects to use with the default of 13K store_avg_size.
>
> Yes, or actually 15000 (the limit is twice what Squid
> estimates based on
> store_avg_obj_size), assuming you are using a recent version of Squid.
>
> Most Squid 1.X versions had 20KB as default store_avg_obj_size and no
> margin, which was a major pain for some people, as the real
> life average
> object size dended to be 13KB.. Nowdays people usually only get this
> error if they have manually changed store_avg_obj_size to a higher
> value.
>
> > Doesn't squid clean up its old entries when it starts to get full.
>
> It does, but sometimes it has some trouble to keep up if the traffic
> rate is high.
>
> > I would think having less available objects would just make your
> > caching less efficient.
>
> ???
>
> > I guess I just want to know if I reduce the size of the
> store_avg_size
> > paramater the same error message isn't going to come back.
>
> If the error is that the actual average object size is much less then
> yes.
>
> > BTW can I safely delete the store.log file.
>
> Yes, or turn if off completely.
>
> squid.conf:
> cache_store_log none
>
> store.log can be regarded as a debug log. It is useful when you
> experiment with refresh patterns or look into why a page is or isn't
> cached, but it is of wery limited use in day-to-day use.
>
> --
> Henrik Nordstrom
> Spare time Squid hacker
>
Received on Wed Aug 11 1999 - 15:54:20 MDT