On 06/08/13 12:08, Kevin Weiler wrote:
Hi again Ceph devs, I'm trying to deploy ceph using puppet and I'm hoping to add my osds non-sequentially. I spoke with dmick on #ceph about this and we both agreed it doesn't seem possible given the documentation. However, I have an example of a ceph cluster that was deployed using ceph-deploy and it seems to have non-sequential osds. I've pasted the output of "ceph osd tree" and my ceph.conf here: http://pastebin.com/0zXvwWcc Is it possible to add osds non-sequentially? It seems like this deployment is doing it. Cheers!
FWIW, it is possible for force this behaviour by adding a bunch of OSDs and them removing them.
Say you wanted your osds to be numbered as '1XX' for rack 1 and '2XX' for rack 2. You'd all 299 osds and then you'd remove the first 100 (0-99).
However, the moment you'd add a new osd, id 0 would be reused instead of assigning it id 300.
How ceph-deploy managed to do such deployment I don't know. Maybe some earlier version hacked around it to do just what the user wanted? Seems farfetched to me.
Do you recall which version were you using when you managed to deploy those osds?
-Joao -- Joao Eduardo Luis Software Engineer | http://inktank.com | http://ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com