QUESTIONS ON CLSTRMGR OF HACMP V1.2
ITEM: RTA000068481
My customer tested the behavior of HACMP (V1.2) with Third-party
takeover configuration (2 active nodes and 1 standby node) and
ran into the following problems.
1. Several times, he could not stop gracefully the clstrmgr on
one node out of 3 nodes. He then tried to stop the clstrmgr
with 'forced' option, and succeeded. Right after that, he
tried to start the clstrmgr on the same node again, but
the "topchng.rc" was never called. (clstrmgr was active.)
After testing several cases, he reached the following result.
When he stopped the clstrmgr at first on the primary active
node or on secondary active node, then stopped the clstrmgr
on the standby node, he could successfully stopped clstrmgr
with graceful option on each node. On the contrary, when he
tried to stop clstrmgr at first on the standby node, then
clstrmgr on the primary or secondary active node, he always
ran into the above problem.
Is this order of stopping clstrmgr officially supported?
Or, is there any other possible cause of the above problem?
2. In some cases (3 cases out of 80, in his test scenario), SMIT
on AIX-window abended with core dump while starting clstr-
mgr. He got into the clstrmgr start menu by entering "smitty
clstart" on one AIX-window and hit enter with all default option
(Start:now, BROADCAST message:no, Startup Cluster Lock Services
:no, Startup Cluster SMUX Peer Daemon:no), then, after several
seconds, the smit abnormally ended with core dump created.
The clstrmgr, however, was started normally.
Do you have any suggestion on the cause of this problem?
If you need the core dump, please let me know. I could send
it to you.
The version of AIX and AIX window is 3.2.4 and 1.2.5 each.
ANSWER
I'll address your questions individually:
1) In a third-party takeover configuration, you MUST stop clstrmgr on
one of the servers FIRST. Then you can stop clstrmgr on the other
nodes. You cannot stop clstrmgr on the stndby node first.
2) I don't know the cause of this problem, although I have seen the
same behavior on our systems running HACMP Version 1.2 (and even
HACMP Version 2.1). Because of this, you have two choices:
a) Ignore the core dump since clstrmgr starts correctly. (It's just
SMIT that core dumps, not HACMP; I found this out by looking at
the core dump that we got).
b) Open a PMR with your defect support center. My understanding is
that you have access to RETAIN, which is the tool you use to do
this.
S e a r c h - k e y w o r d s:
HACMP CORE DUMP THIRD-PARTY TAKEOVER
WWQA: ITEM: RTA000068481 ITEM: RTA000068481
Dated: 05/1995 Category: ITSAIHA6000
This HTML file was generated 99/06/24~12:43:25
Comments or suggestions?
Contact us