in investigating a problem with a system() initiated
execution of "pkill -9 dhcpagent" that we require
as part of our shutdown procedure in the nwam daemon,
it occurred to me that we could use process contract
information to do a cleaner sigsend() to all the members
in the contract (see http://blogs.sun.com/lianep/entry/how_does_kill_work).
This would have the added benefit that it would kill any
ENM scripts that were still running (as we spawn those
also). Is there a reason not to pursue that approach perhaps?
Thanks!

Alan


Michael Hunter wrote:
> On Thu, 26 Mar 2009 17:36:19 +0000
> Alan Maguire <Alan.Maguire at Sun.COM> wrote:
>
>   
>> Michael Hunter wrote:
>>     
>>> Can you gather more data on the pkill failing?  It seems to work for me.
>>>
>>>                     mph
>>>   
>>>       
>> haven't had a chance to do much gathering yet as the
>> system is pretty flaky right now, but poking about in the code
>> it occurred to me on stop, net-nwam could use the
>> contract kill function in /lib/svc/share/smf-include.sh, i.e.
>> replacing
>>     
>
> I talked with Renee about this a bit last night and there was a reason
> some of this stuff works as it does.  I'm pulling a blank at the
> moment.  I intended for this conversation to happen on nwam-dev but I
> got the address wrong.  Want to move over there and see what she has to
> say.  I'll look over the rest of this later after I get some code done.
>
>                       mph
>
>   
>> "/usr/bin/pkill -z `smf_zonename` nwamd"
>>
>> with
>>
>> "smf_kill_contract nwamd TERM"
>>
>> The latter will nuke everything in the contract, including
>> dhcpagent. Even better, if we can map easily from nwamd's pid to the
>> contract id (I'll look into this) within nwamd, we could instead
>> do:
>>
>> "sigsend(P_CTID, ctid, TERM)"
>>
>> within nwamd (though we should probably do it
>> in response to the SHUTDOWN event as opposed
>> to recursively in signal handling context. This would
>> avoid having to use system() or start_child() in
>> the signal handling context.
>>
>> One thing I was wondering though - what do we do
>> about the "pkill -9 nwamd" case? I can't see a way
>> of dealing with that one - is it just a case of tough
>> luck if you do it? Thanks!
>>
>> Alan
>>     


Reply via email to