MQSeries.net :: View topic

Twix · Posted: Mon Oct 17, 2022 2:57 pm Post subject:

Hello

We're currently getting ready to use RDQM at our company.
Until now, we've been mainly using single-instance QMGRs (around 400 of them) running on RHEL VMs (1 QMGR per VM).
In the future, we plan on migrating all of those QMGRs onto 3 RDQM-clusters (one cluster for dev, one for test, one for prod).
Each RDQM-cluster will consist of 3 RHEL VMs. Each VM will have some beefy specs (I'm thinking 64 Cores, 128 GB RAM, but we can always crank this up if needed, since they're VMs).

The idea here is:
- Save on licenses in the long-run, since 3 RDQM-clusters should require less licenses than 400 Single-Instance QMGRs
- Gain High-Availability (taking down a single-instance-MQ-VM for Linux-maintenance currently means we have an outage of that QMGR)

My question is:
Is anyone else using a similar setup? I'm asking because most of the documentation and reports I've seen talked about a setup where you have 1 or maybe 2 QMGRs running on a RDQM-cluster. But this seems like a waste of resources and licenses to me, since one or two of the RDQM-VMs would basically be idle most of the time (aside from receiving the data-replication, of course).

Since we're planning on running around 70-150 QMGRs in total on each RDQM-cluster, I'm asking myself: Is this best practice? Is there a reason *NOT* to do it like that? Is there a likely bottleneck that you can see on the horizon with this kind of setup?

After all, just because it seems like a doable solution *TO ME* doesn't mean it should be done.

So I'm wondering if anyone could chime in with their thoughts.

Thanks!