you already clustered on both servers? If yes, if you want to make it up on server B when server A down with out changes you can do this way.
Disable service in Server B for now. So Server A will be activated. When A goes down, you can enable Server B service and make it work.
1 person found this helpful
Hi Taco - This is the age old EPMA Dimension server question. LOL. Unfortunately the only "supported" option for Active-Passive 9Standby) clustering for EPMA Dimension server is to use Oracle Clusterware. We've had clients over the years using MSCS and had success but in the current release, 22.214.171.124.x Oracle Clusterware is the only tool you should try to use to properly offer some level of high availability. HA solutions should always be supported by the software vendor. That being said, even if you use OCW, there will be a session hiccup/loss as the nodes fail over, so it's not active-active (like Essbase, same model).
EPMA Dimension Server itself isn't really a critical component to the EPM stack - yes it's needed for deployments, however in a DR situation, you can still access your deployed applications and data, etc and if you had to spin up a second EPMA Dimension server box that can be done very easily. We've even had a "staged" VM snapshot on standby for a client who wanted a fast fix. We basically stood up EPMA Dimension Server on it's own EPM Instance (EPMA1) on a VM, made sure there was a viable snapshot of it on a regular basis and in the event of a failover of the real VM (not likely these days), they reverted to the snapshot and were online in less than 5 minutes with service restarts. Not ideal and not technically HA, but it seemed to work since they too didn't want to go down the OCW path.
I'm sure you've heard by now that EPMS is going the way of the Dodo and is no longer included with the upcoming on-premise release 11.2 (Due sometime this September per Oracle at Kscope19's announcement). That being said, I'd suggest looking at your product usage and roadmap. DRM limited use will be included for 11.2 when/if you folks want to look at 11.2 in the coming year. Details on migration paths are yet to be released by Oracle (no matter what anyone on the Internet says, it does not exist yet).
Until then, the 126.96.36.199 docs you've probably already read provide the high level overview of the process for Oracle Clusterware/EPMA Dimension Server HA: https://docs.oracle.com/cd/E57185_01/EPMDO/ch02s03s01.html
I know this doesn't answer your question but after 23 years doing this with Hyperion infrastructure you might want to save your time to use on other issues and basically accept OCW or agree as most clients do, to not worry about HA for EPMA Dimension Server.
If you'd like more details on what we did with our client for their VM / separate instance configuration, just let me know!
Thanks for reply, the thing we did is to manually reconfigure logical web address of dimension server (context: /hyperion-bpma-server) to the server B when server A is down, and manually changing the proxy forward destination in OHS configuration file, and it seems working. Like you said, it's not HA, but since EPMA is not really a critical component so the downtime is accepted.
I wonder even if you have a standalone server for dimension server you still need to do what I mentioned above, right?
Spot on - you would still need to do some DNS magic or repointing or you'd have to edit the EPM System registry - not advisable for failover cases. Your method is very close to what we do. I call it the duct tape fix for EPMA Dimension server failover! Luckily for you, when 11.2 drops in September, you can part with EPMA and get DRM in place. LOL
Hi Taco / Eric i - hope all is well!
Once you're good with this question please let us know by marking it answered so others in the community know you're all set.
Any other questions, please don't hesitate to ask!
Sorry for the late reply, I think the environment is ok now. Changing DNS is not my part of job so I can just leave it to the network administrator, but I can at least have my web server A reaching dimension server B, vise versa.
Thanks for your help!