Failure During Rolling Upgrade of a Pre-9.7.0 Cluster

What if a failure occurs in the middle of a rolling upgrade of a pre-9.7.0 cluster?

A user with an existing HANA resource hierarchy should generate an lkbackup archive before upgrading from a pre-9.7.0 version of LifeKeeper to version 9.7.0 or later to allow for rollback if necessary. See the lkbackup topic for more details.

!We don’t recommend using an lkbackup from an earlier release on a newer release if the upgrade failed.

Possible scenario:

1. User has an existing two-node SAP HANA cluster (node1, node2) running a pre-9.7.0 version of LifeKeeper.
2. Upgrade to LifeKeeper 9.7.0 or later on node2.
3. Bring the HANA resource hierarchy in-service on node2.
   1. The HANA quickCheck script updates the info field on both node1 and node2.
4. A failure of the HANA hierarchy on node2 causes LifeKeeper to fail the hierarchy over to node1.
5. The HANA resource on node1 (still running a pre-9.7.0 version of LifeKeeper) cannot re-establish replication to node2 since it cannot get the replication mode from the updated info field structure.

What errors will occur if the resource is brought in-service on a pre-9.7.0 LifeKeeper system after the info field has been updated by the LifeKeeper 9.7.0 upgrade?

The resource will come in-service on the pre-9.7.0 system but will not re-establish replication to the standby system. The following errors will be in the LifeKeeper log on the in-service system:

• INFO:hana:remoteregisterdb:HANA-SPS_HDB00:136175:failed. trace file nameserver_node2.00000.000.trc may contain more error details.
• ERROR:hana:remoteregisterdb:HANA-SPS_HDB00:136258:Attempt to register server node2 as the secondary SAP HANA System Replication site for database HDB00 failed with exit code 127.
• NOTIFY:hana:remoteregisterdb:HANA-SPS_HDB00:136701:END failed remoteregisterdb of HANA-SPS_HDB00 with return value of 1.
  

How can a user roll back to the pre-9.7.0 format of the info field?

Recommended:

• Restore the resource hierarchy from a recent lkbackup archive on each node taken while running a pre-9.7.0 LifeKeeper version. See the lkbackup topic for more details.