Post B18Z8bF4iCHelvAXEu by linear@nya.social
 (DIR) More posts by linear@nya.social
 (DIR) Post #B18Z8aOxpwbGAImv0y by linear@nya.social
       2025-12-11T11:04:00.194Z
       
       0 likes, 0 repeats
       
       i just decommissioned my biggest server machine (i.e. the one real normal server machine i have, my Poweredge R640) from the Proxmox and Ceph clusters, and have rebooted into the FreeBSD installer.this has been an interesting ~year experiment with Proxmox and Ceph, and that experiment is probably not over, but i think i'm going to begin moving a lot of infrastructure onto the one big reliable reasonably-redundant server
       
 (DIR) Post #B18Z8bF4iCHelvAXEu by linear@nya.social
       2025-12-11T11:13:04.564Z
       
       1 likes, 0 repeats
       
       it will probably be a better experience overall even if i do find ceph's extreme self-healing to be very cool and good.the worst thing i can say about it is: cephfs in this environment can't really do well on workloads that are metadata-heavy or involve lots of small random writes. a lot of this is simply a limitation of the situation i was running it in, and to some degree, the nature of distributed filesystems.my constraints a year ago were that i needed a reasonably reliable storage system out of a hodgepodge of spare drives and mediocre PCs and weird stuff. something that would let me trust that my data was safe, would tolerate multiple computers dying, and ideally was good enough to max out at least a gigabit link, and yknow what? it delivered on all those frontsmy needs today have changed