Thanks for the responses John.

--Scott

On Wed, Feb 24, 2016 at 3:07 AM John Spray <[email protected]> wrote:

> On Tue, Feb 23, 2016 at 5:36 PM, Scottix <[email protected]> wrote:
> > I had a weird thing happen when I was testing an upgrade in a dev
> > environment where I have removed an MDS from a machine a while back.
> >
> > I upgraded to 0.94.6 and low and behold the mds daemon started up on the
> > machine again. I know the /var/lib/ceph/mds folder was removed becaues I
> > renamed it /var/lib/ceph/mds-removed and I definitely have restarted this
> > machine several times with mds not starting before.
> >
> > Only thing I noticed was the auth keys were still in play. I am assuming
> the
> > upgrade recreated the folder and found it still had access so it started
> > back up.
>
> I don't see how the upgrade would have recreated auth keys for the
> MDS, unless you mean some external upgrade script rather than just the
> packages on the machine?
>
> > I am guessing we have to add one more step in the removal mds from this
> post
> >
> http://lists.ceph.com/pipermail/ceph-users-ceph.com/2015-January/045649.html
> >
> >  1 Stop the old MDS
> >  2 Run "ceph mds fail 0"
> >  3 Run "ceph auth del mds.<name>"
>
> Yes, good point, folks should also do the "auth del" part.
>
> > I am a little weary of command 2 since there is no clear depiction of
> what 0
> > is. Is this command better since it is more clear "ceph mds rm 0
> mds.<name>"
>
> '0' in this context means rank 0, i.e. whichever active daemon holds
> that rank at the time.  If you have more than one daemon, the you may
> not need to do this; if the daemon you're removing is not currently
> active (i.e. not holding a rank) then you don't actually need to do
> this.
>
> Cheers,
> John
>
> > Is there anything else that could possibly resurrect it?
>
> Nope, not that I can think of.  I actually don't understand how it got
> resurrected in this instance because removing its
> /var/lib/ceph/mds/... directory should have destroyed its auth keys.
>
> John
>
> > Best,
> > Scott
> >
> > _______________________________________________
> > ceph-users mailing list
> > [email protected]
> > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
> >
>
_______________________________________________
ceph-users mailing list
[email protected]
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Reply via email to