Re: Shutting down NFS

Re: Shutting down NFS

[Jesse Pollard]

VDA <VDA@port.imtp.ilyichevsk.odessa.ua>:
> Hello Jesse,
> 
> Tuesday, August 28, 2001, 4:42:48 AM, you wrote:
> JP> On Mon, 27 Aug 2001, VDA wrote:
> >>...
> >>  killall5 -15; sleep 2; killall5 -9:
> >>    1st run - nothing
> >>    2nd run - nfsd dies
> >>    3rd run - lockd/statd die
> >>    (This is strange. Complicates shutdown script)
> 
> JP> You are using 2 second delay, which might be a bit short, but not unreasonable.
> 
> I have tested this not by shutting down my system but by running a
> test script, watching "ps -AH e" after each run.
> After first run of "killall5 -15; sleep 2; killall5 -9" NFS daemons
> DON'T die at all. After second run only nfsd dies. Only third run
> kills lockd and statd. It does not matter how long I wait between
> runs. (however I didn't wait for minutes. Do you want me to try it?)
> Am I supposed to do the same in shutdown script, i.e.
> 
> killall5 -15; sleep 5; killall5 -9; sleep 5
> killall5 -15; sleep 5; killall5 -9; sleep 5
> killall5 -15; sleep 5; killall5 -9 ?
> 
> This looks ugly and total sleep time is 25 sec.
> A better way is to make NFS daemons understand what user wants after
> first call, not a third.

This already looks like overkill :-) Only the first one should be
needed. I can understand that NFSD could disable signal 15, but not
how it can disable 9... The only way I know for that to happen is
if the process is in an uninterruptable sleep for some reason (and
that should only delay signal delivery, not eliminate it).

I'll have to look at the sources  to get more details.

Everything else looks reasonable (even with a two second sleep).

-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil

Any opinions expressed are solely my own.

Re[2]: Shutting down NFS

[VDA]

Hello Jesse,

Wednesday, August 29, 2001, 3:58:07 PM, you wrote:
>> killall5 -15; sleep 5; killall5 -9; sleep 5
>> killall5 -15; sleep 5; killall5 -9; sleep 5
>> killall5 -15; sleep 5; killall5 -9 ?
>> 
>> This looks ugly and total sleep time is 25 sec.
>> A better way is to make NFS daemons understand what user wants after
>> first call, not a third.

JP> This already looks like overkill :-) Only the first one should be
JP> needed. I can understand that NFSD could disable signal 15, but not
JP> how it can disable 9... The only way I know for that to happen is
JP> if the process is in an uninterruptable sleep for some reason (and
JP> that should only delay signal delivery, not eliminate it).

It looks like killall5 bug - "killall -9 nfsd" kills nfsd at once.
Do you know where killall5 source is? There's no killall5 in
util-linux...

Best regards,
VDA
--
mailto:VDA@port.imtp.ilyichevsk.odessa.ua
http://port.imtp.ilyichevsk.odessa.ua/vda/

Re: Re[2]: Shutting down NFS

[Jesse Pollard]

---------  Received message begins Here  ---------

> 
> Hello Jesse,
> 
> Wednesday, August 29, 2001, 3:58:07 PM, you wrote:
> >> killall5 -15; sleep 5; killall5 -9; sleep 5
> >> killall5 -15; sleep 5; killall5 -9; sleep 5
> >> killall5 -15; sleep 5; killall5 -9 ?
> >> 
> >> This looks ugly and total sleep time is 25 sec.
> >> A better way is to make NFS daemons understand what user wants after
> >> first call, not a third.
> 
> JP> This already looks like overkill :-) Only the first one should be
> JP> needed. I can understand that NFSD could disable signal 15, but not
> JP> how it can disable 9... The only way I know for that to happen is
> JP> if the process is in an uninterruptable sleep for some reason (and
> JP> that should only delay signal delivery, not eliminate it).
> 
> It looks like killall5 bug - "killall -9 nfsd" kills nfsd at once.
> Do you know where killall5 source is? There's no killall5 in
> util-linux...

Glad you found it... The killall5 source is part of the sysvinit. I found
it on the Slackware 8 source disk. (I wonder about the "will not kill
processes in current session" thing... does nfsd appear to be in the
same session???? I wouldn't think so, but that would explain why some of
terminations takes so long, and why some signals appear to be disabled)

-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil

Any opinions expressed are solely my own.