High utilization after OES2 channel upgrade to eDirectory 885 FTF6

  • 7007873
  • 11-Feb-2011
  • 27-Apr-2012

Environment

Novell eDirectory 8.8.5 FTF6
Novell Open Enterprise Server 2 (OES 2) Linux Support Pack 2

Situation

- One or more eDirectory servers upgraded from the channel to eDirectory version 8.8.5 FTF6 ( Binary Version: Novell eDirectory 8.8 SP5 v20506.04, RPM Version: novell-NDSserv-8.8.5.6.0.6.1 )

- High utilization on eDirectory server(s) holding master replica of the partition(s) containing the NCP server object(s) of the upgraded servers

- Several objects getting continuously increasing Revision counters without the entry modification timestamps being updated. (n)dsrepair on the master replica holders reporting and fixing the incorrect entry modification timestamps for the same entries over and over again. This results in the following messages in the (n)dsrepair logs

ERROR: Modification time was incorrect, it has been updated.

- Following messages reported in ndstrace log of the upgraded server(s) with LMBR tag enabled

3053136800 LMBR: [2011/02/10 16:50:49.445] DEBUG: ModifyRDN succeeded.

- Following messages reported for several entries over and over again in the (n)dstrace log of the master replica server(s) with the MISC tab enabled

15:41:53 45858540 0000000A Misc: BumpRevisionOnReferences: SrcID [0005E3D7] RefID [000081FA] succeeded

The SrcID refers to the Entry ID of the recently upgraded server(s) on the master replica server(s)

Resolution

Upgrade to the new version of eDirectory 8.8.5 FTF6 in the OES2 SP2 Channel. The new version has the same binary version.  But the rpm versions can be used to verify that the most up to date 8.8.5 FTF6 has been installed.  The fixed rpm versions are the following:

novell-NDSbase-8.8.5.6-0.6.3
novell-NDSserv-8.8.5.6-0.6.3

Note that if multiple servers have been upgraded to the problematic version of eDirectory, *all* updated servers might need to be updated to the RPM versions indicated above before the high utilisation comes back to acceptable levels. Please contact Novell Technical Support if further clarifications are desired.

Additional Information

This underlying cause of this issue is a code change in the eDirectory Limber background process that attempts to continuously rename the NCP server object of the upgraded server(s) to the same name. This ModifyRDN operation is processed by the server holding the master replica of the partition containing the NCP server object(s) in question. This is followed by an update of all entries that reference the newly renamed server object. The number of entries updated depends on the backlinks to the upgraded server(s). Since the server renames are attempted at very quick intervals, a heavy load on the master replica holder is generated.