Re: Squid AsyncIO and LinuxThread

On Mon, 28 Sep 1998, Henrik Nordstrom wrote:

> If your cache is a production machine then adding a automated stack
> trace may help you (I assume that you do not like to manually restart it
> if it crashes on a production machine).
> 1. Create a modified RunCache that starts "gdb squid" with no options
> instead of "squid -sY"
> 2. Create a .gdbinit file in the directory from where you start
> RunCache, containing the following lines (without linenumber:)
> 1: handle SIGUSR1 nostop noprint
> 2: handle SIGUSR2 nostop noprint
> 3: r -CsY
> 4: bt
> 5: quit
> 3. Start Squid using your modified RunCache..
> 4. Stack traces should be seen in squid.out (assuming you kept the
> redirection from original RunCache script)

Thank you Henrik, this would help a lot. BTW, since newer version of squid
(1.2b24 and up) use SIGTRAP and SIGQUIT instead of SIGUSR1 and SIGUSR2, to
debug them line 1 and 2 above should change accordingly, right?

One thing to note, currently I have modified RunCache to something like
this

---
cd /squid
ulimit -c unlimited
while : ; do
        echo "Running: squid -sY $conf >> $logdir/squid.out 2>&1"
        squid -sY $conf >> $logdir/squid.out 2>&1
        /usr/local/bin/pager xxxxxx "squid on proxy1 is dead"
	...
done
---
When squid got an assertion, I can see assertion message in squid.out. 
What strange is that I never see a message from the system to my pager
when squid got an assertion. (/usr/local/bin/pager use to send a message
to my pager). On the other hand, I got a message instantly when manually
shutdown squid with 'squid -k shutdown' and RunCache happily restart squid
shortly after that. 
I'm not sure if running squid under gdb will behave the same, if so (is
this a bug?) I think we don't get any stack trace. I'll try and report
back again.
Regards,
--
aet