Need to setup HA on WIPDS using EGO

[SASWayne]

I'm new to setup HA on services. Our environment consist of three nodes, one for each metadata, compute and middle tier. We have WIPDS setup on all three nodes where the instance on compute is primary. Now, we're planning to setup HA on WIPDS where it failovers to metadata node.
I believe the LSF daemons are setup only on the compute node as the bhosts command returns only that node.
is it possible to setup EGO to failover WIPDS to metadata node where LSF daemons are not installed?

[JuanS_OCS]

Hi there,

excellent question, it is technically possible but highly not recommended, as I would expect it won’t be officially not supported and you will be basically on your own from that moment on.

The trick would be to generate a script that would act as service wrapper (meaning, it has to contain all parameters of the service interface) on the compute, were EGO runs. The wrapper would be connecting via SSH to the appropriate nodes to replicate the calls.

Then, let me stress this, it is not a supported design. If I would be you I would purchase the LSF license for your metadata and web nodes, should you want to achieve WIP HA with EGO.

Nonetheless, you can also achieve a supported version of WIP HA without EGO.

See this paper:
http://support.sas.com/resources/papers/Managing-WIP-DataServerforHA.pdf

I hope it helps,
Best regards,
Juan

[SASWayne]

@JuanS_OCS
Thanks! We have a requirement to do it through EGO.
Do you think it is a good idea to add our metadata server to the cluster (close it for accepting job) and then trying EGO would work?

[JuanS_OCS · Accepted solution]

Hi @SASWayne !

when you have a license for any other node (more cores), being the meta or web one, you can do something as what you mentioned, yes.

Still, I would contact SAS Tech Support, as they might have an opinion or best practice you can follow for your requirement.