February 26, 2020 at 9:01 am
I would recommend adding either a disk or file share to the 3-node cluster giving (by default) 4 votes. Then - I would set the node weight of the 3rd node (reporting) to zero, effectively taking it out of quorum. Basically, you now have a 2 node cluster and quorum is based on the 2 actual nodes and your witness.
With that setup you can take down reporting at any time and not affect the cluster. You can also take down reporting - and lose another server and still be operational with a single server and your witness (disk/file).
Now - with that said...what is the purpose of setting up an AG for HA only in the same building? Do you have separate SANs for each node or is all storage coming from the same SAN? If the same SAN then your redundancy isn't really redundant...lose the SAN and you lose all servers in the cluster anyways.
So...why not just utilize a standard FCI 2-node cluster and add a 3rd node (setting the node weight to 0 and adding a disk/file share witness) in an AG just for reporting? Works exactly the same...except you don't have to duplicate the data across 3 different sets of volumes on the SAN. The only downside (that I am aware of) would be a slightly longer fail over time...if it even takes longer.
Same SAN , I know it's not perfect , but it's someone who's salary is triple that of mine doing the design (consultant and i'm new in the business, he's been here for quite a while)
not a chance of going to a data centre and getting a nice failover solution when we just spent 3/4 of a million on hardware on his advice.
to be honest you guys are giving me the confidence to push back and say, " whats the point in a 3 node cluster on one server" - however the performance of the shycronos replication is a factor
MVDBA
February 26, 2020 at 9:26 am
It does all come down to what you need Mike at the end of the day.
Sounds like you want something to provide local HA should a server have a problem, hence the 2 sync nodes and a reporting box to take away reporting load.
Obviously single point of failure is the SAN on all aspects here.
Is AG the way to go, maybe, maybe not. I would of counteracted it with saying what about a potential traditional FCI, shared storage etc, that way you've saves a load of disk space already, got the local HA you need. Then something like replication (transactional or snapshot) or log shipping for reporting.
As always more than one way to skin a cat to get to where you need. Appreciate your between a rock and a hard place at the moment as it someone else's design.
Just remember to get your read-only routing and application intents rights on the AG setup and connection strings for reporting otherwise it's going to be going to the read-write replica.
Viewing 2 posts - 16 through 16 (of 16 total)
You must be logged in to reply to this topic. Login to reply