Re: Squid 1.1.22 dies on USR1 kill

On Wed, 22 Jul 1998, Bruce Hodge wrote:

> We have a 20Gig cache running squid 1.1.22 under Solaris 2.55.
> The cache is showing that is running with about 200 + Meg of Memory
> There seems to be plenty of Swap free (about 200Mb)
> But when we cycle the logs on a daily basis it crashes with the
> following error messages :-
> ftpInitialize: fork: (12) Not enough space
> Failed to fork() for ftpget
> and then dies.
> There should be enough space the partion is 30gig?
> And there is plenty of swap?

Bruce

Note that this is a memory problem and has nothing to do with the amount
of cache_swap area you have available.

I noticed the same effect on one of my Squids ... the solution appears to
be to reduce Squid's memory usage to the point where it is _very_ light on
swap-space usage (ie: only a few %).

I had once upon a time used cache_mem set at 1/3 of real available RAM,
but even this seemed too high on this particular machine, which is
periodically quite heavily loaded.

The symptom appeared when I noticed that a considerable amount of swap was
in use; sending a SIGUSR1 while the machine was in this state caused
errors (in my case) for restarting the DNS servers - with a similar error
message to the one you saw, and then Squid collapsed and was restarted by
RunCache.

Reducing cache_mem further has apparently alleviated the problem (I'm down
to 1/5 of real RAM ... but I plan to tweak this until I have a reasonable
compromise).

I hope this helps.

John.

----------------------------------------------------
John V Cougar | Voice: 1800 065 744
Cache Manager |-----------------------------
Telstra Internet | E-Mail: cougar@telstra.net
----------------------------------------------------
Received on Tue Jul 21 1998 - 16:34:17 MDT