On 04/22/2015 03:47 PM, SCHAER Frederic wrote: > Hi, > > I've seen and read a few things about ceph-crush-location and I think that's what I need. > What I need (want to try) is : a way to have SSDs in non-dedicated hosts, but also to put those SSDs in a dedicated ceph root. > > From what I read, using ceph-crush-location, I could add a hostname with a SSD suffix in case the tool is called against a SSD... thing is : I must make sure this is a SSD, and this is where coding and experimenting comes. > > Hence, I'd like to know if someone would have an already working implementation that would detect if the OSD is a SSD, and if so, append a string to the hostname ? > I'm for instance wondering when this tool is called, if the OSD is already mounted (or should have been...), what happens at boot ... > > I know I can get the OSD mountpoint using something like that on a running OSD : > #ceph --format xml --admin-daemon /var/run/ceph/ceph-osd.0.asok config get osd_data|sed -e 's|./osd_data.*||;s|.*osd_data.||' > /var/lib/ceph/osd/ceph-0 > > I know I can find out if this is a disk or a SSD using for instance this : > [root@ceph0 ~]# cat /sys/block/sdy/queue/rotational > 0 > [root@ceph0 ~]# cat /sys/block/sda/queue/rotational > 1 > > So I just have to associate the mountpoint with the device... provided OSD is mounted when the tool is called. > Anyone willing to share experience with ceph-crush-location ? > Something like this? https://gist.github.com/wido/5d26d88366e28e25e23d I've used that a couple of times. Wido > Thanks > > > > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > -- Wido den Hollander 42on B.V. Ceph trainer and consultant Phone: +31 (0)20 700 9902 Skype: contact42on _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |