We are testing a configuration with a two-node TruCluster v5.1A (with
patches)
installation, where Oracle 9i database is running on one node. I registered
this as
a cluster application and set it with 3 restarts and 180 seconds timeout.
We let the
systems run and found that several times the ckpt (checkpoint) process
(apparently)
died without leaving messages in the Oracle logs. This caused the database
to be
restarted. We also had the dbw0 (database writer) process (apparently) go
away
when a database import was interrupted with Ctrl C. This also caused the
database
to be restarted.
The question is whether any of you have tested Oracle 9i database in a
TruCluster
v5.1A configuration and have observed similar problems. If so, did you find
a
resolution?
Second issue is that when the cluster management process "declared" the
database
application not working right and tried to either restart on that node or
relocate it
to the other node, our shutdown is with the IMMEDIATE qualifier, yet the 180
second
wait was not long enough if someone was in the middle of an import, a long
query, or
something else with an active connection to the database.
This question is whether you have determined the shutdown method and
some
reasonable time for the timeout.
If anyone has tested Oracle 9i database on TruCluster v5.1A (or v5.1 for
that
matter in a cluster), I would appreciate your insight on these or other
issues you
dealt with.
Thanks in advance and I will summarize,
Jenny Butler
jbutler_at_utmem.edu
Received on Tue Jan 28 2003 - 19:51:01 NZDT