2024-09-12T12:09:38.327Z | <Ken Carlile> Anyone have any experience with reimaging a host while keeping the OSDs intact using cephadm? I've found an article on IBM's site that talks about the process, but since I'm not using IBM's spin on things, just a straight cephadm/ceph orch setup, I'm curious if there's anything else I should be paying attention to. <https://www.ibm.com/docs/en/storage-ceph/7.1?topic=wrn-replacing-node-by-reinstalling-operating-system-using-ceph-osd-disks-from-failed-node> |
2024-09-12T12:13:54.323Z | <Eugen Block> It's really straight forward, I wrote a short [article](http://heiterbiswolkig.blogs.nde.ag/2021/02/08/cephadm-reusing-osds-on-reinstalled-server/) for cephadm, and it still works well. I had this case just a few months back on a customer cluster which had a couple of other steps because it was a former SUSE Enterprise Storage customer. But the basic steps should still work. |
2024-09-12T12:15:39.914Z | <Ken Carlile> Cool, thanks! I had run across your article as well, but since it was from 2021, I wasn't sure if things had changed at all. |
2024-09-12T12:16:14.651Z | <Eugen Block> I haven't followed the steps from my blog article though, I might have to update it, because I haven't had to create any symlinks lately... I'll need to check on that when I have more time. But in general, you just need to prepare the host to be reachable by the mgr again (ssh config, passwordless sudo, etc.), and when the host is reachable, `ceph orch osd activate <host>` should just do it (for me it did). |
2024-09-12T12:16:44.975Z | <Ken Carlile> nice. |
2024-09-12T12:17:06.596Z | <Ken Carlile> Now I just need Dell to ship me the cables for the boot drives... |
2024-09-12T12:17:46.138Z | <Eugen Block> Hold on, the command was from the top of my head, it's actually:
`ceph cephadm osd activate <host>`
See the [docs](https://docs.ceph.com/en/latest/cephadm/services/osd/#activate-existing-osds) for the details. |
2024-09-12T12:19:14.788Z | <Ken Carlile> cute... I haven't run across a `ceph cephadm` command before... yet another variant to try and remember. |
2024-09-12T12:20:36.921Z | <Eugen Block> It's really helpful, for example to login all your hosts to a custom registry, or get the current ssh-config and a couple more |
2024-09-12T16:33:49.746Z | <verdurin> Restarting the OSD containers seems to be working.
Many thanks. |
2024-09-12T16:43:53.560Z | <Eugen Block> Good to hear, thanks for the update. |
2024-09-12T21:15:32.659Z | <Ken Carlile> well, that didn't work as expected... |
2024-09-12T21:45:31.977Z | <Ken Carlile> I apparently didn't thoroughly cleanse our kickstart file of all the stuff that destroys extant LVMs and whatnot... |
2024-09-12T21:45:44.989Z | <Ken Carlile> whoopsiedoodle. Good thing there's no data on this thing yet |