1

I have a bunch of Always On clusters that I manage and am really comfortable with them. I have one client who is running an Always On cluster with 2 replicas. They have 4 AG's that each have their own distinct listener and have more than 10 databases per AG. At their request (mostly because I could not think of any reason to not do it), we have 2 AG's primary on each replica. It looks like this in the Always On Dashboard.

Server1 AG1 - Primary, AG2 - Primary, AG3 - Secondary, AG4 - Secondary

Server2 AG1 - Secondary, AG2 - Secondary, AG3 - Primary, AG4 - Primary

Everything is running well. We have spread the workload over 2 servers this way and are in effect Active\Active. I am wondering if anyone else has this configuration and if there are any issues that anyone has run into because of it. I have looked all over the place for known issues and cannot find them. If you have this same type of Always On configuration, what has been your experience with it?

The only negative I can find so far is that syncing logins and jobs can get more complicated but that is it so far.

Thanks!

2
  • 1
    If a server goes offline (perhaps for patching, or maybe because it caught on fire), does your DR plan require you be online running on a single node? Will your hardware support running on a single node?
    – AMtwo
    Aug 11 '21 at 2:57
  • If a server goes offline then all AG's do run on one node. The hardware will support it.
    – LoriB
    Aug 11 '21 at 14:22
0

You'll need to automate some kind of "load balancing" otherwise, if there is an issue with one one the node during the night, all AGs will run on the same server. Unless you want to fail them over manually every time, you'll need some kind of automatisation.

Be carefull when building that as you don't want your AG to failover all the time back and forth.

The other thing to consider is the server sizing.

Let's say each of your AGs uses 20% of the ressources, then each server will be used at aroung 40% and in case of a downtime in one of the server, then the remaining server will be used to around 80%. (All good)

The thing is that if ever each AG goes from 20% to let's say 40%, you will not notice any issue as each server will use around 80% of their ressources (so no problem), but if one of the node die, you'll suddently need 160% of the remaining server's ressources which will cause performance issue until the broken node is fix.

Those are the 2 main "cons" I can think of.

1
  • Thanks! Those are the only cons I can think of too. The hardware will support all AG's running on one node but the CPU utilization may be a bit high until we can get the second node up again.
    – LoriB
    Aug 11 '21 at 14:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.