[squid-users] Too much assertion failed: comm.cc:178: "fd_table[conn->fd].halfClosedReader != NULL"

[Amos Jeffries]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 16/01/2015 1:50 a.m., Yuri Voinov wrote:
> 
> Hi gents,
> 
> Sometimes I have too much assertions:
> 
> 2015/01/08 09:40:39 kid1| assertion failed: comm.cc:178: 
> "fd_table[conn->fd].halfClosedReader != NULL"
> 
> Core dump not produced but Squid restarts. So, I can't get stack
> trace.
> 
> When I set hals_closed_clients to on, I've got much more
> assertions:
> 
> 2015/01/15 18:41:31 kid1| assertion failed: comm.cc:1823:
> "isOpen(fd) && !commHasHalfClosedMonitor(fd)"
> 
> What can be wrong? This is very annoying behaviour. Also memory
> cache is often cold after restarts and hit ratio is degrades.
> 
> This looks similar adaptation bug 4057, but it is closed in 3.4.11.
> This behaviour depends half-closed connections, but how?
> 
> And how to troubleshoot?

A) running Squid under GDB. There are instructions in
<http://wiki.squid-cache.org/SquidFaq/BugReporting> on how to do that
even if you are debugging a production server that needs no downtime.

OR,

B) finding out how to enable core dumps in your OS. The assertion
handling calls abort() which will generate a core file if the OS
allows them to be created.

Amos
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (MingW32)

iQEbBAEBAgAGBQJUt71IAAoJELJo5wb/XPRjVJEH+OPwizmwffDGN4nrKW/OwfI3
X4Pu15cXgie+6WWBsVBuz7FHlOv846oL6B2oSGHWTcAy2iR9E1+BZc/BtcIR+hZ+
vxjSmisDT7h9NvynVelrA4yqw2hrZeVVSNyuaaLk1ks2tvTYb6ms7EldGN2TKDUg
Q1aXgofE4p+5mXrZB8vEnJrU4CrXp/8pGJ6+YM4tQhDcI+fvcy4rSi7lTxbbxXnL
jsSLSnq9rNXKtBB3Ig6gVF3nQhwhl0IpGw571ORAuRQ7uvDC5d/fT/uNOpWfOFA5
4H1VR8RY/IddxEkhqdbnkYn6fGzk6qyOV3ZQKQi8p2mcyuNGfpAKbD9ZC+v28A==
=uSQr
-----END PGP SIGNATURE-----