Hi Sage, On Fri, May 9, 2014 at 12:32 AM, Sage Weil <sage at inktank.com> wrote: > > > On Fri, 9 May 2014, Indra Pramana wrote: > > > Hi Sage, > > Thanks for your reply! > > > > Actually what I want is to replace the journal disk only, while I want > to keep the OSD FS > > intact. > > > > I have 4 OSDs on a node of 4 spinning disks (sdb, sdc, sdd, sde) and 2 > SSDs (sdf and sdg) > > > > osd.28 on /dev/sdb, journal on /dev/sdf1 > > osd.29 on /dev/sdc, journal on /dev/sdf2 > > osd.30 on /dev/sdd, journal on /dev/sdg1 > > osd.31 on /dev/sde, journal on /dev/sdg2 > > > > I want to replace the SSDs (sdf and sdg) without losing the data on the > OSD. I believe > > ceph-disk prepare will destroy the data on the OSD? > > Right. > > Your procedure looks about right, assumin gyou are using ceph.conf to > indicate what the data and journal paths are: > That's the problem, we don't. :) We used ceph-deploy to prepare the OSDs so I believe we are using the so-called "udev magic" that you mentioned earlier? The journal is mounted on this file (or folder?): /var/lib/ceph/osd/ceph-XXX/journal, which in turns is a symbolic link to this file: lrwxrwxrwx 1 root root 58 Apr 6 23:45 journal -> /dev/disk/by-partuuid/3ff2c20c-6b58-41d3-9e7e-b3ec63c62c2f which in turns is a symbolic link to the journal partition on the device: lrwxrwxrwx 1 root root 10 May 3 23:48 3ff2c20c-6b58-41d3-9e7e-b3ec63c62c2f -> ../../sdf1 Since we don't use ceph.conf to indicate the data and journal paths, how can I recreate the journal partitions? Looking forward to your reply, thank you. Cheers. > > > > > - set noout > > - stop the osds > > - flush the journal > > - replace journal SSDs > - recreate journal partitions > - update ceph.conf to reflect new journal device names > > - recreate the journal (for the existing osds) > > - start the osds > > - unset noout > > sage > > > > > > > Is it possible? > > > > Looking forward to your reply, thank you. > > > > Cheers. > > > > > > > > On Fri, May 9, 2014 at 12:12 AM, Sage Weil <sage at inktank.com> wrote: > > Hi Indra, > > > > The simplest way to do the fs and journal creation is to use the > ceph-disk > > tool: > > > > ceph-disk prepare FSDDISK JOURNALDISK > > > > For example, > > > > ceph-disk prepare /dev/sdb # put fs and journal on same > disk, or > > ceph-disk prepare /dev/sdb /dev/sdc # fs on sdb, journal on (a > new part on) sdc > > > > It will create the partitions, label them, and then create the (by > > default, XFS) fs and initialize the journal. After that, udev > magic will > > take care of all the mounting and starting of daemons for you. > > > > sage > > > > > > On Fri, 9 May 2014, Indra Pramana wrote: > > > > > Hi Sage, > > > Sorry to chip you in, do you have any comments on this? Since I > noted you advised > > Tim Snider on > > > similar situation before. :) > > > > > > http://www.spinics.net/lists/ceph-users/msg05142.html > > > > > > Looking forward to your reply, thank you. > > > > > > Cheers. > > > > > > > > > > > > On Wed, May 7, 2014 at 11:31 AM, Indra Pramana <indra at sg.or.id> > wrote: > > > Hi Craig and all, > > > > > > I checked S?bastien Han's blog post, it seems that the way how the > journal was mounted > > is a > > > bit different, is it because the article was based on older version of > Ceph? > > > > > > ==== > > > $ sudo mount /dev/sdc /journal > > > > > > $ ceph-osd -i 2 --mkjournal > > > 2012-08-16 13:29:58.735095 7ff0c4b58780 -1 created new journal > /journal/journal for > > > object store /srv/ceph/osd2 > > > > > > $ sudo service ceph start osd.2 > > > === osd.2 === > > > Starting Ceph osd.2 on ceph03... > > > starting osd.2 at :/0 osd_data /srv/ceph/osd2 /journal/journal > > > ==== > > > > > > From what I can see, on all my OSD nodes in my Ceph cluster, the > journal is mounted on > > > this folder (or file?) instead of /journal: > > > > > > /var/lib/ceph/osd/ceph-X/journal > > > > > > which in turns is a symbolic link to this file: > > > > > > lrwxrwxrwx 1 root root 58 Apr 6 23:45 journal -> > > > /dev/disk/by-partuuid/3ff2c20c-6b58-41d3-9e7e-b3ec63c62c2f > > > > > > which in turns is a symbolic link to the journal partition on the > device: > > > > > > lrwxrwxrwx 1 root root 10 May 3 23:48 > 3ff2c20c-6b58-41d3-9e7e-b3ec63c62c2f -> > > ../../sdf1 > > > > > > I am using one SSD for journals for multiple OSDs within a node. Any > advise on the > > > correct way how to mount and create the journal? Do I need to mount > first, or create > > > first? Because I supposed that we need to mount to the partition > *after* the journal is > > > created? > > > > > > I am using latest stable version of Dumpling (v0.67.7). > > > > > > Any advice is greatly appreciated. > > > > > > Thank you. > > > > > > > > > > > > On Wed, May 7, 2014 at 1:40 AM, Craig Lewis <clewis at centraldesktop.com> > wrote: > > > On 5/6/14 03:34 , Gandalf Corvotempesta wrote: > > > > > > Hi to all, > > > I would like to replace a disk used as journal (one partition for each > OSD) > > > > > > Which is the safest method to do so? > > > _______________________________________________ > > > ceph-users mailing list > > > ceph-users at lists.ceph.com > > > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > > > > > > > > I haven't tried this yet, but I imagine that the process is similar to > moving your > > > journal from the spinning disk to an SSD. > > > > > > S?bastien Han had a blog post about this: > > > > http://www.sebastien-han.fr/blog/2012/08/17/ceph-storage-node-maintenance/ > > > > > > > > > > > > -- > > > > > > Craig Lewis > > > Senior Systems Engineer > > > Office +1.714.602.1309 > > > Email clewis at centraldesktop.com > > > > > > Central Desktop. Work together in ways you never thought possible. > > > Connect with us Website | Twitter | Facebook | LinkedIn | > Blog > > > > > > > > > _______________________________________________ > > > ceph-users mailing list > > > ceph-users at lists.ceph.com > > > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > > > > > > > > > > > > > > > > > > > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.ceph.com/pipermail/ceph-users-ceph.com/attachments/20140509/55a8686f/attachment.htm>
 |