On Fri, Nov 29, 2013 at 9:46 AM, Gautam Saxena <gsaxena@xxxxxxxxxxx> wrote:
> I've got ceph up and running on a 3-node centos 6.4 cluster. However, after
> I
>
> a) set the cluster to nout as follows: ceph osd set noout
> b) rebooted 1 node
> c) logged into that 1 node, I tried to do: service ceph start osd.12
>
> but it returned with error message:
>
> /etc/init.d/ceph: osd.12 not found (/etc/ceph/ceph.conf defines ,
> /var/lib/ceph defines )
>
> Now, it *IS* true that the /etc/ceph/ceph.conf does NOT make any reference
> to osd.12.
>
> Am I supposed to *manually* update the master ceph.conf file to put each and
> ever osd reference into (osd.1 through osd 14 in my case) and then copy the
> ceph.conf file to each node?
Yes. ceph-deploy does not create specific sections for daemons. You
can create those and then push the config everywhere though with:
ceph-deploy config push {nodes}
I would have thought that when I add a new osd
> (using ceph-deploy) that it would *automatically* update either a) the
> master ceph.conf file on my admin machine (where ceph-deploy runs) or least
> it would update the ceph.conf file on the node that contains the osd that's
> being added.
Since ceph-deploy doesn't even create them (the daemon sections) then
no, it doesn't update those.
>
> It feels as if ceph-deploy will allow you to add osds, but this addition
> only updates the "in-memory" definitions of ceph.conf, and that after server
> reboot, it looses the in-memory definitions and attempts to re-read info
> from ceph.conf. Is this correct?
That is not entirely correct if I understand you correctly. If you
don't have daemons specified in your configuration file it means that
you would need to specify them in the command line.
For example:
sudo /etc/init.d/ceph start osd.0
If so, is there a way when using
> ceph-deploy to get it to automtically create the necessary entries for the
> ceph.conf file? If so, how? If not, then am we supposed to both a) use
> ceph-deploy to add osds and then b) manually edit ceph.conf? If so, isn't
> that dangerous/error prone?
A lot of questions here :)
* There is currently no way to make ceph-deploy to automatically
create daemon sections in ceph.conf
* Yes, you need to add those yourself and then push the config (using
the example I added above)
* That may be error prone, but it is also complexity in ceph-deploy
that is not needed for someone to get started with a Ceph cluster. For
granular operations of ceph clusters (including configuration updates)
I would recommend using a configuration management tool (Chef, Puppet,
Ansible, etc...)
That is also documented in "Why feature X is not implemented" in the
documentation:
http://ceph.com/ceph-deploy/docs/#why-is-feature-x-not-implemented
>
> Or am I missing something fundamental?
>
> (I'm using ceph version 0.72.1).
>
> _______________________________________________
> ceph-users mailing list
> ceph-users@xxxxxxxxxxxxxx
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
