Best Practices
Failover scenarios
Suggest changes
PDFs
Uniform access setup
In a uniform access configuration, the fibre channel SAN is stretched across both sites. The ESX hosts at both sites could access both copies of the data sets. During normal operation, the ESX host running the HANA system is accessing the local copy of the data based on proximity settings in the initiator group configuration. Each ESX host has an active optimized path to the local copy and an active non-optimized path to the mirrored copy.
Normal operation
During normal operation the HANA system reads and writes from/to the local copy based on the active optimized path from ESX host ESX-1. With each backup operation, SnapCenter detects the current primary site for the replication relationship and executes the backup operations against the primary site. The Snapshots are replicated to the mirrored copy and are available at both sites. A SnapCenter restore operation would be executed at the primary site.
Storage failure
If the storage system at site 1 fails, the HANA systems access the mirrored copy at site 2 and continues operation. The primary site switches to the secondary site and SnapCenter now executes backup and restore operations at the new primary site.
Site failure
In case of a site failure, the HANA VM as well as SnapCenter and the SnapCenter for VMware plugin VM will fail over to the ESX host at the secondary site using vSphere HA. The HANA database needs to get started and will then access the mirrored copy at the second site. The primary site switches to the secondary site and SnapCenter now executes backup and restore operations at the new primary site.
Non-uniform access setup
In a non-uniform access configuration, the fibre channel SAN is not stretched across both sites. Each ESX host at each site can only access the local copy of the data sets.
Normal operation
During normal operation the HANA system reads and writes from/to the local copy. With each backup operation, SnapCenter detects the current primary site for the replication relationship and executes the backup operations against the primary site. The Snapshots are replicated to the mirrored copy and are available at both sites. A SnapCenter restore operation would be executed at the primary site.
Storage failure
In case of a storage failure, the HANA VM as well as SnapCenter and the SnapCenter for VMware plugin VM will fail over to the ESX host at the secondary site using vSphere HA. The HANA database needs to get started and will then access the mirrored copy at the second site. The primary site switches to the secondary site and SnapCenter now executes backup and restore operations at the new primary site.
Site failure
Same as storage failure.
Relocation of HANA VM or primary site
If the HANA VM is relocated to the other ESX host and the primary site of the storage remains the same, a restore operation with SnapCenter will fail. Since SnapCenter uses the primary site to execute restore operations, the clone will be created at the left side, while the HANA VM runs on the right side. Since there is no data path between the sites, SnapCenter will not be copy the data.
As a workaround you need to make sure, that the relocation of VM and primary side is done together, or you need to failover the primary site before the restore operation with SnapCenter.
