Regards managers,
Seems like my previus posting arose some thoughts in the Alpha sysadmin
community judging all the answers I got. And that's good. I just want to
clarify things:
1. Altering an export does not necessitate a stop in NFS services in a
normal environment. So why should this be necessary in a High
Availability configuration ? This is bad HA implementation if you ask
me. It should not be necessary to see the infamous 'server not
responding' message on clients just because you alter a exportlist on
the server. I can understand and accept that there are some tasks that
will require a full stop and restart but this is not one of them !
2. Let's say I configured my start/shutdownscripts correctly. Anyhow,
this is not sufficient to guarantee a clean shutdown/restart, I believe.
Consider this:
I use 'lsof' to detect open files in the service I am about to change.
'lsof' reports everything OK, so I enter 'asemgr'. I do the changes and
commit them. If the shutdown script is run and setup correctly, it
probably will take care of any residue processes that 'lsof' didn't
catch in the first place. BUT, there have to be a window of time between
the handling of these 'stray' processes and the effective ASE service
shutdown, so there still can be processes accessing the service even
after the shutdown script is executed (e.g NetWorker doing a automatic
restart of a previously failed backup or something else undetectable by
'lsof' or similar tools). This should definitely be handled otherwise. I
still think it's a bug.
Any comments ?
--
******************************************************************
* Knut Hellebų | DAMN GOOD COFFEE !! *
* Norsk Hydro a.s | (and hot too) *
* Phone: +47 55 996870, Fax: +47 55 996495 | *
* Cellular Phone: +47 93092402 | *
* E-mail: Knut.Hellebo_at_nho.hydro.com | Dale Cooper, FBI *
******************************************************************
Received on Thu Oct 02 1997 - 13:14:39 NZST