I had a lot of problems with this on one of my servers running Networker 6
after upgrading from 5. I had to add my own set of lock files to my pre and
post commands.
At the time I had a pretty good idea of the sequence of events in the pre and
post command processing. One thing I remember, and your log shows this, is
that when the worklist is empty the post command is run. My problem was that
the worklist would empty multiple times during my backup.
I ran this script in parallel with my backup to watch what was going on
(mohawk_ora is the savegrp, magic is the server, oraback* are my lock files):
#!/bin/ksh
# nsrspy.ksh
#
# Periodically list the nsr processes to try to figure out why the
# savepnpc postcmd is being run in the middle of the backups.
#
typeset -i i
i=1
while (( i < 360 ))
do
date
ps -ef | grep [n]sr | grep -v nsrspy
ps -ef | grep [s]ave
(( i = i + 1 ))
echo
ls -l /nsr/tmp/mohawk_ora.??? 2>/dev/null
ls -l /nsr/res/oraback* 2>/dev/null
echo
echo "
show work list
print type: NSR group;
name: mohawk_ora" | nsradmin -s magic -i -
echo
echo
sleep 60
done
exit
In your case, it seems that the worklist is being checked before preclntsave
gets a chance to put anything on it. I don't remember why the *clntsave
processes aren't being "spyed" on. I think I remember that savegrp starts a
preclntsave that runs the precmd and then sets the lock file so the next
saveset that gets started doesn't run the precmd again. The first preclntsave
watches the worklist and when it sees the worklist is empty it lets
postclntsave run.
My problems were on an older client sending the backup to an even older server
with slow network and an overtaxed library. Networker's internal timeouts
didn't seem to be long enough to deal with my environment.
Are your parallelisms (client,server,etc.) set high enough that your savesets
on this machine overlap? Looking back, that may have been my problem. My
settings may have been such that I would get a couple of savesets started, but
not all of them. When these completed they would free up enough resouces to
allow some more to start. That wasn't acceptable in my case since I was
backing up an Oracle database with tablespace files spread across my savesets.
If you "spy" on your backup, look at the save processes. There will be one for
each saveset. If the number running drops to 0 before they've all been run,
your postcmd will be run.
>From: "Hicks, Adrian" <Adrian.Hicks AT BARTTER.COM DOT AU>
>Date: 3/15/04 4:55PM
>Subject: Re: precmd & pstcmd RES file
>
>More info.
>There seems to be a time out at the same time that the "pstcmd" is
>started as below. The DB shutdown is completed at 23:56 and I launch a
>mailx for the "pstcmd" and it was generated at that time.
>
>Does this help ??
>Thanks
>Adrian
>
>"savepnpc.log"
>
>03/15/04 23:56:12 preclntsave: All command(s) ran successfully.
>03/15/04 23:57:12 pstclntsave: Time out condition occurred.
>03/15/04 23:57:19 pstclntsave: All command(s) ran successfully.
>03/16/04 05:58:20 pstclntsave: All savesets on the worklist are done.
>03/16/04 05:58:20 pstclntsave: Exited.
>
>-----Original Message-----
>From: Legato NetWorker discussion [mailto:NETWORKER AT LISTMAIL.TEMPLE DOT EDU]
>On Behalf Of Davina Treiber
>Sent: Tuesday, 16 March 2004 4:32 AM
>To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
>Subject: Re: [Networker] precmd & pstcmd RES file
>
>Hicks, Adrian wrote:
>> Hi,
>>
>> I have been having trouble for months regarding the "pstcmd" running
>> straight after the "precmd" script has run. This has only started
>> since I upgraded from Networker 6.1.1 to 7.1 on the server only,
>> client servers are still on 6.1.1 and they run fine. My OS is Tru64
>> ver.5.1. At one stage I had some old /nsr/tmp/*.tmp file which I have
>> cleaned to solve some RPC errors but it didn't fix this error
>>mentioned above.
>
> That's a new one on me, but I have done more than a little work with
>savepnpc over the years. What does the savepnpc.log say?
>
---
Jim McAlarney Huber Engineered Materials
phone: 404-949-5505 4401 Northside Parkway
fax: 404-949-1305 Suite 600
jmcalarney AT huber DOT com Atlanta, GA 30327
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|