Sun, 5 Dec 1999 21:54:45 -0500
|
On Sat, 04 Dec 1999 09:07:14 EST, Bill Verity <[log in to unmask]> said:
> What tools or techniques to you use to monitor listserv's status on AIX?
>
> On VM, we watched the reader file queue. If it got large, someone was called.
> This was automated so it worked without depending on frail humans :-)
>
> We are just now moving to the AIX platform. I came in this morning
> and found the server had stopped running at 00:30. There were no
> errors in the log. It just stopped. Have to catch this earlier
> somehow.
A good trick under AIX is to get /etc/init to (a) launch listserv
at system IPL time and (b) restart it if it terminates. We have this
in /etc/inittab:
listserv:2:respawn:/home/listserv/go bg
which re-launches it if it exits (due to 'respawn'). The
way we roll the logs every night is with a 'cron' job that
just does a 'kill -TERM' of the lsv process, and a little bit
of magic in the 'go.user' code - since that's a shell script,
and it's being run when the server starts, we *know* it's a good
time to roll all the logs.
I second the comments about checking process limits - we had problems
with the lsv process going *poof* until we changed the 'Soft DATA segment'
limit to -1 (go to 'Change/Show Characteristics of a User' in SMIT).
This is particularly important if you've raised the values of the
FIOC_* variables in go.user.
Also, use 'lsps -a' and make sure you have enough paging space allocated.
/Valdis
|
|
|