SAP HANA clustered using HPE ServiceGuard supported by plugin for HANA?

[mdiver]

Does the plugin for HANA support a HANA cluster (SLES15SP5) clustered using HPE Serviceguard for Linux?
If so, what is the approach? Only use the shared IP? Or include all nodes seperately?

Thanks,
Mike

[PetrM]

Hi Mike,

I assume you are referring to SAP HANA System Replication and not Scale-Out System? If this is the case, you will not be able to back up from replicas, and you should only focus on managing failover. You can find more information on this page. Technically, this configuration should work without any issues. However, please note that we officially support this configuration for Linux on Power only.

Thanks!

[mdiver]

Hi Petr.

Thanks for your reply.

With HPE Serviceguard on top of system replication we get a shared common cluster IP for both nodes failing over between them. Failover is nearly transparent.
The plan was to setup the plugin backup using the shared cluster IP to be able to backup no matter were the active DB is currently running.
From what you say, it seems unlikely this would work?

Here an intro to how Serviceguard works with HANA system replication:
https://www.hpe.com/h22228/video-galler ... 995c/video

Thanks, Mike.

[PetrM]

Hello,

In fact, it's vice versa: technically, it should work just fine. Veeam B&R will see the cluster as one entity and the failover operations will be transparent. However, the configuration with HPE Servicguard is not supported officially, we never tested it, and we won't troubleshoot technical issues if they are related to the specifics of this setup.

I think that the official support of HPE Serviceguard would be a good improvement for our future releases, but I cannot share ETA.

Thanks!

[mdiver]

Hi Petr.

Thanks for your reply. It took us some time to setup the environment.
Now we testet in a PoC with a two node HANA cluster on Serviceguard.
The backup from the primary node using the shared IP ran fine and at full speed.

After having switched to the secondary node using HPE Serviceguard, the backup failed with an error regarding the fingerprint of the underlying Linux. Quite reasonable as the system switched without VBR knowing it.

So, we tried to add both systems as trusted hosts using the same FQDN/IP using "Add unknown hosts to the list manually" in Linux host authentication.
Unfortunately, this does not seem to work. VBR seems to parse only the first item in the list of systems when having more than one fingerprint for the same FQDN/IP.

I guess we're out of luck here and have to backup the nodes independently.
It would be great to have Serviceguard supported as we see this quite often with HANA.

Thanks,
Mike

[PetrM]

Hi Mike,

Many thanks for the information provided! Indeed, the fact that the nodes share the same IP leads to this fingerprint issue. It seems that the only way forward is to process the nodes separately, but perhaps it would be possible to script the preferred node selection somehow. Your request is noted, but I still have no information about the ETA.

Thanks!