Monitor/Manager/Metadata Nodes? How many? Where?

[Seed]

Hello.

I have the following:

5 hosts total
3 data/compute nodes (4x10TB OSD (CEPH), NVMe VMs, 64GB RAM 8c/16c) machines 10GBe
2 4c/8t 16GB RAM nodes with a local SSD also on 10GBe but no local ceph disks

1. Can I just put manager/metadata/monitor on the 2 smaller machines and be good or does there need to be 3?
2. Should I just put manager/metadata/monitor on all 5 nodes?
3. Put manager/metadata/monitor on the 3 compute nodes. I might just retire the smaller 2 nodes then.

I'm mostly trying to understand what the overhead really is with these 3 services if they run on the storage and computer nodes or stand alones.

Thanks!

 

[Alwin]

Our precondition section might help to answer your questions.
https://pve.proxmox.com/pve-docs/chapter-pveceph.html#_precondition