On Wed, Mar 9, 2022 at 11:10 AM ravi k <kottapar(a)gmail.com> wrote:
Hello,
We are creating a self-hosted engine deployment and have come up with a draft design. I
thought I'll get your thoughts on improving it. It is still a test setup and so we can
make changes to make it resilient.
We have four hosts, host01..04. I did the self-hosted engine deployment on the first node
which created a dc, cluster01 and a storage domain hosted_storage. I added host02 also as
a self-hosted engine host to cluster01.
Now the questions :-)
1. It is recommended not to use this SD hosted_storage for regular VMs.
Indeed.
So I'll create another SD dc_sd01. Should I use this dc_sd01 and
cluster01 when creating regular VMs?
Yes.
What's the best practice?
These are two slightly different questions, but the answer is yes anyway.
If you are going to have more hosts, please note that it's not
recommended to have more than 8 hosted-engine hosts. So if you'll
still want to keep them all in the same cluster, some of them will be
HE hosts and some not - this might be slightly confusing, depending on
your use case.
2. It is a bit confusing to get my head around this concept of
running regular VMs on these self-hosted engine hosts. Can I just run regular VMs in these
hosts and they'll run fine?
Yes.
3. Please do suggest any other recommendations from experience in
terms of designing the clusters, storage domains etc. It'll help as it is a new setup
and we have the scope to make changes.
You can try searching the list archives for previous discussions about
topology/architecture that people had over the years.
You might want to check also this somewhat-old but still mostly
relevant doc, which is for RHV, but probably applies 99% to oVirt as
well:
https://www.redhat.com/en/resources/best-practice-rhv-technology-detail
Good luck and best regards,
--
Didi