re: Squid 2.2.2-STABLE2 RPMs from Andrew Herrick

Just got the chance to upgrade my ageing squid system to the 2,2 RPMs built by
the kind Mr. Herrick.

A few minor things emerged:

a) Not all possible run states have a corresponding rc file, viz:

rpm -ql squid | grep etc/rc

/etc/rc.d/init.d/squid
/etc/rc.d/rc0.d/K07squid
/etc/rc.d/rc3.d/S87squid
/etc/rc.d/rc6.d/K07squid

For completeness, there ought to be:

/etc/rc.d/rc1.d/K07squid
/etc/rc.d/rc6.d/K07squid

as well - I agree it's really really minor, but there you go.

b) The install sequence left squid not running.

     For which I mean it failed to link

     /etc/rc.d/rc3.d/S87squid -> /etc/rc.d/init.d/squid

     and thus make it "runnable"

     This after an rpm upgrade from the really ancient squid RPM which came with
 RH5.1.

     1999/03/23 19:31:37| Starting Squid Cache version 1.1.21 for
i686-pc-linux-gnu....

     I think this was partially because the format of the init script is not
entirely compatible with
     "chkconfig".

     To this end I added a few extra lines to the front of the init script, and
invoked chkconfig to add
     the links.

.....

[root@intranet /root]# cat /etc/rc.d/init.d/squid
#! /bin/sh
#
# chkconfig: 345 87 07
# description: Squid is the proxy cache that allows http access over the \
# firewall.
# processname: squid
# pidfile: /var/run/squid.pid
#

.......

c) The squid init script itself seems to cope badly when squid crashes
FATALLY.

As it transpired I got:

FATAL: You've run out of swap file numbers.

before I fixed some parameters.

As a consequence, squid died without removing /var/run/squid.pid

As a further consequence, /etc/rc.d/init.d/squid start failed to restart squid
because squid.pid still existed

Therefore added:

     rm -f /var/run/squid.pid

to the end of the stop sequence as a workround.

I can't help feeling that actually the squid binary itself ought to always
remove the pid file on the way down,
irrespective of the failure mode. This is for gurus to argue over at length.....

=============================================================================

Re the FATAL error, I had this in squid.conf:

cache_mem 32 MB
store_avg_object_size 13 KB

before I dropped the avg_size to 8 KB

cache_mem 32 MB
store_avg_object_size 8 KB

which seems to have fixed things for the moment, but no doubt I'll be tweaking
other things in the near future.

The squid process is up to some 50MB Resident Set Size after 5 hours uptime.

Ted Rule,

Flextech Television
Received on Tue May 04 1999 - 09:04:20 MDT