On Fri, Mar 14, 2008 at 10:20:21PM +0000, "Stephen" wrote:
> Hi,
>
> When my cache is busy, if I issue a SQUID -K RECONFIGURE then Squid very
> often crashes with:
>
> FATAL: Too many queued url_rewriter requests (54 on 12)
>
> This seems only to happen when the cache is busy. Once the FATAL has
> occurred, Squid needs to be restarted manually.
I've run into this too. There is a heuristic in helper.c:
if (hlp->stats.queue_size > hlp->n_running * 2)
fatalf("Too many queued %s requests (%d on %d)", hlp->id_name, ...
If *2 is not suitable for your environment, increase it, or take out the
fatalf() entirely (although then squid loses the ability to react to all
of your rewriters going out to lunch)
> Changing the number of url_rewriters does not seem to make any difference.
> [...]
> I am using SquidGuard 1.3 as the url_rewriter. All DBs are in binary
> format, so startup time is not long.
Sounds like the startup time is long enough that you're accumulating
enough queued requests for squid to kill itself. Increasing the number
of url_rewriters might not help, as it'll take longer to fire up more
processes.
> Also, issuing the reconfigure when the cache is not being used or is
> under light load is never a problem.
Right, because the queue doesn't build up as quickly and swamp the
rewriters.
--Emil
