MID cluster topology best practices?

[jMarshal]

Hi all - we're looking to start using some ITOM features and in preparation, I need to improve our MID environment. We are currently just using 1 MID for LDAP integration purposes and a couple daily data export jobs. The business is concerned with service performance and continuity. Is there a best practice topology when setting up for both load-balancing and fail-over scenarios?

My initial thought is a load-balanced cluster for each separate capability/use and then replicate that in an over-arching fail-over cluster, making the footprint of the MID environment at least twice the size as it would be, if just load-balanced.

...but I'm wondering if a load-balanced cluster will also fail-over with it's standard configuration and operational behaviour? -- or will the entire cluster fail, if one machine fails? -- if it keeps running, what happens when you get down to one? Will it continue to operate as a "cluster" or will it now fail, as there is only one MID remaining?

For instance, which is better (in theory) -- one cluster of 4 mids, all doing load-balancing (and fail-over) for the same capability/task -- or two clusters of 2 mids each, with each of the clusters being load-balanced between themselves, but replicated to provide fail-over to the other?

Perhaps I'm over-thinking this, but I'd rather have a good plan, than get part way through and deal with a misstep. Thanks in advance! 

[Daniel Borkowi1 · Accepted solution]

Hi @jMarshal ,

first of all I would recommend to separate integration MIDs from Discovery/Service Mapping MIDs. The best practice is to place these servers as close as possible to the devices you want to discover. This ensures that you don't make lot of holes in your firewalls (you only need one port open 443 and not almost every port) and you reduce the load on your firewalls. Next point load balancing cluster are also failover clusters. When on MID in cluster went down the over one will do the job alone. I always recommend to use load balancing cluster. I would built up a cluster of 2 MIDs in you main networks zones. But start small and increase you compute power if needed.

Hope that helps.

Greets Daniel

[jMarshal]

Thanks @Daniel Borkowi1 -- this does help. The fact that load balancing clusters will still operate as failover clusters, is what I was wondering. 

[Daniel Borkowi1]

@jMarshal, if my answer is all what you need, then please mark it as correct to close this thread. 

Greets Daniel