Anti-colocation rule required for SAP Hana Cost Optimized Scenario as of SLE 15 SP6

SUSE Linux Enterprise Server for SAP Applications 15 SP6
SUSE Linux Enterprise High Availability Extension 15 SP6
SAP HANA System Replication Scale-Up with Cost Optimized Scenario

The system was configured following the SUSE Best Practice Guide: SAP HANA System Replication Scale-Up - Cost Optimized Scenario 

After upgrading to SLES for SAP Applications 15 SP6, once the former primary node joins the cluster, the following will happen:
 

• The IP resource (rsc_ip_HA1_HDB10) for the replicated HANA instance will move to the former primary node
• The non-replicated QAS HANA (rsc_SAP_QAS_HDB20) will start on the secondary node
• The replicated HANA (rsc_SAPHana_HA1_HDB10) will be promoted on the former primary node.
  It will not join the existing Primary node.

Below is an example of the failed situation:

2024-11-12T11:52:03.482188+01:00 suse02 corosync[105782]:   [TOTEM ] A new membership (10.156.61.172:3167) was formed. Members joined: 1
2024-11-12T11:52:58.591552+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Move       rsc_ip_HA1_HDB10                    (     suse02 -> suse01 )
2024-11-12T11:52:58.591618+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Start      rsc_SAPHanaTopology_HA1_HDB10:1     (                 suse01 )
2024-11-12T11:52:58.591666+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Start      rsc_SAPHana_HA1_HDB10:1             (                 suse01 )
2024-11-12T11:52:58.591711+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Start      rsc_SAP_QAS_HDB20                   (                 suse02 )
...
2024-11-12T11:53:28.723126+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Start      rsc_SAPHanaTopology_HA1_HDB10:1     (                 suse01 )
2024-11-12T11:53:28.723194+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Demote     rsc_SAPHana_HA1_HDB10:0             ( Master -> Slave suse02 )
2024-11-12T11:53:28.723242+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Promote    rsc_SAPHana_HA1_HDB10:1             ( Stopped -> Master suse01 )
2024-11-12T11:53:31.777210+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Promote    rsc_SAPHana_HA1_HDB10:1             ( Stopped -> Master suse01 )
...
2024-11-12T11:56:04.369606+01:00 suse02 pacemaker-schedulerd[105983]:  notice: Actions: Promote    rsc_SAPHana_HA1_HDB10:1             (   Slave -> Master suse01 )

To get the desired cluster behavior, the following extra location rule is needed: 

location loc_QAS_never_on_promoted_node rsc_SAP_QAS_HDB20 \
       rule -inf: hana_ha1_clone_state eq PROMOTED

Note: Please adjust the above rule to match the existing configuration. 

Below is an example of the desired behavior after adding above location rule: 

2024-11-12T12:18:13.102237+01:00 suse01 SAPHana(rsc_SAPHana_HA1_HDB10)[4684]: INFO: LPA: landscape: DOWN, LPA: register ==> try to register
2024-11-12T12:18:13.116844+01:00 suse01 SAPHana(rsc_SAPHana_HA1_HDB10)[4684]: INFO: DEC: ANOTHER HANA IS AVAILABLE ==> LETS REGISTER
2024-11-12T12:18:13.414632+01:00 suse01 SAPHana(rsc_SAPHana_HA1_HDB10)[4684]: INFO: ACT: REGISTER: hdbnsutil -sr_register --remoteHost=suse02 --remoteInstance=22 --replicationMode=sync --operationMode=logreplay --name=WDF
2024-11-12T12:18:17.628201+01:00 suse01 SAPHana(rsc_SAPHana_HA1_HDB10)[4684]: INFO: ACT: Register successful

Starting with SUSE Linux Enterprise High Availability Extension 15 SP6, pacemaker has been improved to handle anti-colocation more thoroughly for resource allocation. This means, a more specific rule setting for the location of the SAPInstance resource is required now.

Prior SP6, it worked without this additional location rule because pacemaker did not take anti-colocation rules sufficiently into consideration upon calculating scores for resource allocation. This happened to prevent SAPInstance from being assigned to the secondary site where SAPHana is already promoted. 

Note: Adding the new constraint rule also works well on the former Service Packs. It should not cause any issues.

Kindly open a new case with SUSE Technical Support/Consultation in case of doubts to discuss the recommended changes.