Whee, almost lost ~20% of the data on this filesystem two days before I intended to finally upgrade my offsite backup host to be able to back it up completely. Murphy is one hell of an asshole.
Show this thread
-
Narrowly avoid disaster with the "re-create RAID array with the same parameters, assume clean, fsck and cross fingers" approach. I had some corrupted parity from a transient glitch doing major damage to the FS.
Show this thread
I'm getting tired of RAID and traditional filesystems. Stuff doesn't scale. The backup host is going to be moving to Ceph (single-host Ceph, but hey, why not) and if that works out I might migrate my home NAS to that too.
-
-
Why would you run a single host Ceph cluster? To gain experience? IMO mdraid is more reliable and easier to manage. Good luck!
-
mdraid does not scale. Rebuilds take forever, there's no intelligence, write-intent bitmap is nice but coarse, filesystem integrity is a bitch, no checksums, no ability to have disparate-sized disks. I have a 20TB md-raid array now and have had too many close calls.
- Show replies
New conversation -
-
-
I've been thinking of running a modestly sized SAN box with LizardFS or XtreemFS.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Ceph’s a bit of a nightmare when it comes to perf degradation/recovery, depending on how you set things up. My company’s moving off of it after years of trying to get it stable and resilient. Different scale, though, and usage pattern; I’ve seen people have v good luck w/ it. GL.
-
You planning on monolithic RBD’s, CephFS for multi-client, or something else?
- Show replies
New conversation -
-
-
cephfs, or using the object store directly?
-
CephFS and RBD. I might play around with performance pros/cons and see what I want to use for what data.
End of conversation
New conversation -
-
-
This Tweet is unavailable.
-
Ceph *is* in the kernel. The kernel side is upstream. And the other side is userspace. ZFS is still an out-of-tree module :P
- Show replies
-
-
-
I migrated to Ceph half a year ago, and it has been working great. It does not really like SMR drives, but those with an SSD cache in front behave very well as most writes are sequential-ish later. I still have to deal with an in-place btrfs + mdadm raid6 shrink though
-
I just upgraded all my drives from 4T to 8T, so my plan is going to be to hold off on resizing the array until I decide I'm switching to Ceph, then put OSDs on the free space, move the data over, and finally re-provision one drive at a time to be fully an OSD.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.