named stop not stopping

[Lisa Besko]

Doug Barton wrote:
> Lisa Besko wrote:
>> We had an issue with the /etc/rc.d/named script this morning where it
>> looped/hung in the wait_for_pids subroutine.  We run a job to restart
>> named which calls the /etc/rc.d/named script with a stop command.  For
>> some reason named did not stop properly and the process continued to
>> list PIDs from the wait_for_pids subroutine forever.  Has anyone ever
>> encountered this before?
>>
>> We are running 7.2 stable and bind 9.4.3-P3 on an AMD system.
> 
> First, please verify that you have the latest version of
> /etc/rc.d/named installed.
> 
> Second, this could happen if the call in that script to rndc
> succeeded, but for some reason rndc was not able to actually kill the
> binary. In that case there should be something in the logs that
> indicates what the problem is.
> 

I'm running a current supported version of bind.  It's the one that was 
included with 7.2 stable and has been patched.

The only thing I see in my logs is the icmp unreach message but that 
does not start until after the named stop command was issued.


kernel: Limiting icmp unreach response from 259 to 200 packets/sec


Currently we changed the way we run our restart script to  do a kill -9 
if all else fails.  The /etc/rc.d/named script uses a kill -TERM if the 
original rcnd stop fails.  In this case that wasn't enough.

Thanks for the input.

-- 
Lisa Besko