Re: [PVE-User] pve-user Digest, Vol 160, Issue 13

[Christoph Weber <Christoph.Weber@xpecto.com>]

Thanks for your advice, Ralf and Eneko ,

@Ralph
> please read the documentation about removing/replacing a Node carefully
> - you may not reinstall it with the same IP and/or Name as this will crash your
> whole cluster!

This is my bigges fear ...

Thanks for pointing me to the manual - there is in fact an detailed explanation which files to keep before reinstalling the node under "Re-installing a cluster node". 

This is excactly what I was looking for, with detailed step-by-step walktrough.  Though I'm not sure if I will give it a try - looks a bit risky to mess things up in one of the many manual steps.

> For the boot disks we use 2 or 3 mirrored zfs disks to be sure...

Sounds like a good idea in retrospect ;-) 
Until now I just was under the impression that the system disk does not contain very important data and can easily be replaced ...
 
@Eneko Lacunza
> Do you have Ceph OSD journal/DB/WALs on system disk?


No, we have everything belonging to one OSDs on the corresponding SSD drive.

> Moving OSDs from node3 to node6 would trigger data movement, but I'd go

Will it? I thought it might just relocate the ODSs location in the crush map from node3 to node6 when I shut down the prodve3, remove the disks and reinsert them in node 6? At least that was my impression in a thread here on the mailinglist a few weeks ago.

> node3, "pvecm delnode" it, reinstall with new system disk and rejoin cluster.

This seems to be the best solution. 

Maybe we will just discontinue the node, as it is getting near its planned lifetime ...