I would consider doing it host-by-host wise, as you should always be able to handle the complete loss of a node. This would be much faster in the end as you save a lot of time not migrating data back and forth. However this can lead to problems if your cluster is not configured according to the hardware performance given.
-- Martin Verges Managing director Mobile: +49 174 9335695 E-Mail: martin.ver...@croit.io Chat: https://t.me/MartinVerges croit GmbH, Freseniusstr. 31h, 81247 Munich CEO: Martin Verges - VAT-ID: DE310638492 Com. register: Amtsgericht Munich HRB 231263 Web: https://croit.io YouTube: https://goo.gl/PGE1Bx Am Fr., 15. Nov. 2019 um 20:46 Uhr schrieb Janne Johansson < icepic...@gmail.com>: > Den fre 15 nov. 2019 kl 19:40 skrev Mike Cave <mc...@uvic.ca>: > >> So would you recommend doing an entire node at the same time or per-osd? >> > > You should be able to do it per-OSD (or per-disk in case you run more than > one OSD per disk), to minimize data movement over the network, letting > other OSDs on the same host take a bit of the load while re-making the > disks one by one. You can use "ceph osd reweight <number> 0.0" to make the > particular OSD release its data but still claim it supplies $crush-weight > to the host, meaning the other disks will have to take its data more or > less. > Moving data between disks in the same host usually goes lots faster than > over the network to other hosts. > > -- > May the most significant bit of your life be positive. > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com