Hotfix install on cluster with load sharing

[Rui_Pereira1]

I pretend do install an hotfix R80.10 on a cluster environment with 2 GW.

When it is a ClusterXL HA, you install in the backup node, CU failover to backup node, and let it work 1 or 2 days. If there is no issue, the hotfix is installed on the main node.

With this approach there are 1 or 2 days the versions mismatch on the Cluster, so in this period failovers are avoided and if a failover happens, there could be some service degradation.

When the cluster is in load-sharing mode, I suppose the 1-2 days of different versions is something to be avoided at all cost, correct?

In this case, how can we proceed?

Update both at the same time? In case of issues, the hotfix must be uninstalled or use a previous snapshot.

Which one will be updated first? The pivot or the other?

The load share can be changed from 30/70 to 0/100 or 100/0, but this changes the database, which is something that I like to avoid.

Or should we convert to ClusterXL HA, do the "normal" update, and the convert again to load-sharing?

Which is the best procedure for hotfix update/install on load-sharing clusters?

[Kaspars_Zibarts · Accepted solution]

I would definitely do both at the same time. As in not leaving 2 days in-between. Then uninstall hotfix or restore snapshot as you suggested yourself. Which is your current version and which hotfix are you installing?

[Rui_Pereira1]

Being a cluster LB with 2 GW, leaving 2 days between will lead to 30% of load with an older HF, and 70% of the load with the new HF, which in case of issues can be difficult to troubleshoot.

Current version: R80.10 Take 421 with HF 56.

To install: HF 169.

[Kaspars_Zibarts]

Exactly. That's why doing both is much better idea