Hi all,
Using filestore, we have some clusters were we put some journals of
regular osds(hdd) together with eg. cache or metadata osd on one SSD.
Even with the OS too on the OSD, this gave us better performance than
with journals on disk.
Now using bluestore, i was thinking if it is possible to have a
bluestore OSD on an SSD, together with the block.db/block.wal of HDD
osds ? Something like this for the SSD:
Number Start End Size File system Name Flags
1 1049kB 106MB 105MB xfs ceph data
2 106MB 150GB 150GB ceph block
3 150GB 151GB 1074MB ceph block.db
4 151GB 152GB 604MB ceph block.wal
Using ceph-deploy/ceph-disk , this does not seem possible at the moment.
Adding the db/wal partitions is not a problem, but having the OSD to
share the disk is:
- ceph-disk does not accept partitions, it needs full disks to make the
xfs and block partitions
- it always needs to have the first two partitions
- it will take all the space left of the disk for the OSD block partition.
I probably could hack something in by resizing the partitions, like
above, but I'd rather not :)
Will such kind of feature be possible, or is this just a bad idea with
bluestore?
Thank you very much!
Kenneth
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
