Index Storage Clustering

[Andrew_Tankersl]

We are soon going to migrate our EV servers to new servers. Most likely we will be spinning up a primary server, with a second server standing by for HA.

We have the ability to have our indexes (stored on a SAN) presented to both servers. The idea being that the indexes would already be available on the standby server if needed in a fail-over scenario.

Have any of you tried such a setup?

[AndrewB]

i dont really think it's possible with SAN but if you're doing it somehow, you'll want to be careful about LUN corruption. what i would do is setup all the wwpn's adhead of time but not mount the LUN on the standby server. to me that's the safest way and it really doesn't take any time to mount once all the backend has been configured.

[Andrew_Tankersl]

Our plan was to use EMC's Symclone to copy the indexes over to the secondary server. What I understand about Symclone is that the LUN would be presented\available unless there was an issue and then we would ask our storage team to present the LUN to the secondary server

[LCT]

I don't think there is a problem having the indexes presented to the two servers however, you should really seperate the Indexes on a separate SAN or LUNs for the HA server (this would work as a DR too) as AndrewB rightly pointed out that if the Indexes get corrupted then you will have major problems regardless of which server it is presented to.

Just a matter of interest how are you presenting the Vault Store Partitions to the HA server? Commonly people would use a third party tool such as double take or EvNearSync to implement similar solutions to yours. Just have to weigh out pros and cons for each tool. Again this wouldalso be ideal for DR too and not just HA.

Hope that helps.

[AndrewB · Accepted solution]

in that case you're copying or replicating the LUN (to somewhere else) and presenting the replica to the second EV server. i interpreted your original post differently, hense my first comments.

i'm not an expert with EMC's Symclone but conceptually, i see no issue with what you describe.

[Andrew_Tankersl]

Thansk Andrew - I'll update this post  when we decide if we're going to go this route