Ancestors

Written by Rachel on 2025-01-16 at 01:15

Cluster rebuild planning:

I can punt the node idea by just running two ceph VMS on the fileserver/proxmox (plenty of space to shove drives)

When I finally build out a decent third node I can move the disks/osds over there

Only one problem.

The server rack is downstairs and I am really cozy under warm blankets on the sofa

=> More informations about this toot | More toots from rachel@transitory.social

Toot

Written by Rachel on 2025-01-16 at 21:05

Actually I don't need to shuffle drives if I run one OSD per rook/ceph node, or mix drive sizes a bit

In no particular order plan is:

• Attempt to lower r630/r730 power draw to same levels (remove HBA, one CPU)
• Test fileserver power with HBA removed to evaluate the viability of another Ryzen-based compute node (which would get the GPU)
• Actually test out the odroid. Godamn if they had made a few different design choices these could make perfect nodes, I really wish I could get a viable 1u case for them

If I can convince myself that 2.5gbs is enough I could maybe run a few odroid nodes?

=> More informations about this toot | More toots from rachel@transitory.social

Descendants

Written by Michael on 2025-01-16 at 21:46

@rachel Do the HBA cards really draw so much power that removing them reduces the total power draw by that much?

=> More informations about this toot | More toots from mmeier@social.mei-home.net