On Fri, 11 Mar 2022, Abe E wrote:
Has anyone setup hype converged gluster (3Nodes) and then added more
after while maintaining access to the engine?
I have added additional self-hosted-engine hosts to a cluster in 4.3 and
it worked fine. I don't know if 4.4 is more strict about that or not, as
I moved away from self-hosted for 4.4
An oversight on my end was 2 fold, Engine gluster being on engine
nodes
and new nodes requiring their own cluster due to different CPU type.
You don't _have_ to use the default CPU type. If they are just newer
models of the same type, you can just add them to the existing cluster as
the old CPU type. I have a mix of Skylake and Cascade Lake CPUs in my
Skylake cluster, and it works great. I guess you could have issues if
your original cluster is too old to have "Secure" CPU types, and your new
CPUs only have "Secure" CPUs, or if your old cluster is Intel and new one
is AMD.
So basically I am trying to see if I can setup a new cluster for my
other nodes that require it while trying to give them ability to run the
engine and ofcourse because they arent part of the engine cluster, we
all know how that goes. Has anyone dealt with this or worked around it,
any advices?
From a gluster standpoint, this should work fine, but I suspect oVirt
won't like having self-hosted engine hosts in different clusters.
Unless you plan on entirely removing the original hosts in the future, I'm
not sure this should be much of an issue anyway. While the idea of being
able to run the engine on any host is nice, it's also a bit overkill.
Three hosts should be sufficient for redundancy.
If you just want to have the engine running on the newer hosts because
they may be around longer, I would just make the new hosts part of the
existing cluster (assuming they're close enough), migrate everything to
the new hosts, then remove the old hosts and rebuild them as
non-self-hosted hosts and put them in a separate cluster. Then once
that's done, you can upgrade the cluster CPU type on the new machines.