RE: rac cluster crashes

  • From: "Crisler, Jon" <Jon.Crisler@xxxxxxx>
  • To: <adar666@xxxxxxxxxxxx>
  • Date: Fri, 24 Apr 2009 11:50:59 -0400

If you went back on 10.2.0.4 and ran the cluster verification utility,
it probably would have found this issue.  CVU can be downloaded and
installed separately from any patch or RAC install media.

 

________________________________

From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Yechiel Adar
Sent: Thursday, April 23, 2009 5:33 AM
Cc: oracle-l@xxxxxxxxxxxxx
Subject: Re: rac cluster crashes

 

My co-workers found the problem late last night.
It seems that the network connection has a name like: private.

When installing the RAC they wrote: Private  (capital first latter).
It worked fine in 10.2.0.3.
Oracle did more severe error checking and they failed in 10.2.0.4 with
the last CPU because the first letter was 'P' instead of 'p'.
No log or error message to let you know.

BTW - They first tried to change the name of the network connection but
it was restored after boot.
They got from Oracle support a command to change the name in Oracle
configuration file, so it will match.



Adar Yechiel
Rechovot, Israel



Yechiel Adar wrote: 

Where can I find the VIP trace file.
My co-worker did an upgrade last night, on windows 64 bit from oracle
10.2.0.3. to 10.2.0.4 + the last CPU.
I got called at midnight because the VIP did not work after the upgrade.

When he went to bed, they changed the tnsnames to use physical address
instead of vip and that is the way it is working now.



Adar Yechiel
Rechovot, Israel



Mark Strickland wrote: 

This sounds kinda similar to what happened to us after upgrading from
10.1.0.3 to 10.1.0.5.  From my posting in 2006 about our
Upgrade-from-Hell
"//www.freelists.org/post/oracle-l/Upgrade-from-92-to-102-NONRac-Ba
sic-Steps,2":

[Oracle Support] was able to identify the cause of the problem in the
VIP trace file.  It was occasionally timing out while checking the
 
default gateway.  The timeout threshold was 2 seconds and the engineer
had us change that to 10. 


Perhaps this will help.

-Mark

Other related posts: