XI 3.1 odd behavior

All,
I need some ideas. We have clustered 2 nodes windows 3.1 installation which was upgraded using the upgrade package form 3.0 sometimes in March. We did not apply yet any fix packs or SP2. CMS is an Oracle 10g DB. All worked fine until recently. Now every time I am searching for a user in CMC

  1. It can not find the user
  2. It shows the group hierarchy but if the group is selected users are not visible in the right panel for the selected group.
  3. If I am looking for the users by selecting the “Users List” right panel displays nothing but it shows that we have multiple pages on the top right corner. Selecting the next page again shows nothing.
  4. After one or two selection CMC crushes.
    Going to CCM it shows that SIA is running but logging through Managed Servers not possible. Gives me an error CMS is not available. Event Viewer on the apps server shows that SIA lost the connection to CMS. BO logs files shows the following

“SvcMgr_8488_2009_09_13_14_58_49_352_trace.log”

.\ServerInfo.cpp:1660: TraceLog message 1
2009/09/13 14:58:49.352|>>|A| | 8488|5548| |||||||||||||||assert failure: (.\ServerInfo.cpp:1660). (FALSE : Error locating registry settings for Server Intelligence Agent).
.\ServerInfo.cpp:1660: TraceLog message 2
2009/09/13 14:58:52.680|>>|A| | 8488|5548| |||||||||||||||assert failure: (.\ServerInfo.cpp:1660). (FALSE : Error locating registry settings for Server Intelligence Agent).

“cms_4488_2009_09_13_14_52_37_805_trace.log”

.\Aggregation.cpp:1575: TraceLog message 1
2009/09/13 14:52:37.805|>>|A| | 4488|3236| |||||||||||||||assert failure: (.\Aggregation.cpp:1575). (k != aliasToUser.end() : Unknown alias returned from auth plugin’s GetAttributes().
).

In order to restore the connection I have to restart SIA. Any suggestions or ideas will be welcomed.


icotler :us: (BOB member since 2002-08-19)

Are you using AD or LDAP authentication?

It looks like either one of the users/groups are corrupt. Try to re-add the AD groups in the CMS. Are you using Alias for your users?

Also since you are running a cluster it could be a good test to shut down one node and test and then reverse it to test the other node, just to check that there isn’t any issue with the nodes.


J0sh :australia: (BOB member since 2006-10-09)

AD authentication.
Are you suggesting than one corrupted group can affect all of the users in all the groups?


icotler :us: (BOB member since 2002-08-19)

[Moderator Note: Moved to CMC - Jansi]


Jansi :india: (BOB member since 2008-05-12)