Well , here's another interesting one.
My Linux kernel (Redhat 6.2) seems
dislike "maintaining" squid :-) :-)
This happened yesterday night :
Jul 4 07:49:25 dmz1 kernel: VM: killing process squid
Jul 4 07:49:25 dmz1 squid[762]: Squid Parent: child process 28204 exited due to
signal 9
Jul 4 07:49:28 dmz1 squid[762]: Squid Parent: child process 22281 started
Well, I have plenty of swap, but for the moment I am not so
much concerned with the cause of this problem.
The thing is that I had a discussion with Duane in the past
considering Squid crash recovery.
I had a problem with squid crashing on standard "squid -k shutdown",however
in that case the parent would nog recognize the special condition
and restart Squid anyway.
This would make it sometimes difficult to >really< stop squid.
Duane developed a patch for this so that the parent would exit
too on SIGKILL.
This discussion assumes however that a "kill -9" would be human
operator issued, so to speak.
Now if the kernel can do that too, the validity of killing the
parent comes under discussion !
I made humble plea in the past to the squid developers , to
support for instance a new directive in the conf file as in :
squid_sigrecv_action restart.
Well restart would be a default, and the other value would
be "stop" for instance.
These would be directives for the parent so to speak.
I think it could be usefull , especially for people maintaining
cache hierarchies, or testing development configurations of SQUID.
To keep track of what is going on
Of course this setup could be elaborated as in :
squid_sigsegv_action ...
squid_sigkill_action ...
squid_siganyother_action ...
Marc.
Received on Wed Jul 04 2001 - 07:32:43 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:01:00 MST