Unfortunately we are not in control of the VMs using this pool, so something like "sync -> stop VM -> incremental sync -> start VM on new pool" would be extremely complicated. I _think_ it's possible to misuse a cache tier to do this (add a cache tier, remove the underlying tier, add a new pool and remove cache tier), but that's a hack at best.
So before we go even considering this - will there be any significant gains from this? When we increased the PG numbers it had a very positive effect on the cluster, but with only 1/3 of the drives I am worried there will be too much contention on the OSDs. I've already seen a higher CPU usage and while some latency metrics went down thanks to the new Intel drives, other metrics went up of course, so I'm not sure how it will perform in the real life... Jan > On 01 Sep 2015, at 18:08, Robert LeBlanc <[email protected]> wrote: > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > We are in a situation where we need to decrease PG for a pool as well. One > thought is to live migrate with block copy to a new pool with the right > number of PGs and then once they are all moved delete the old pool. We don't > have a lot of data in that pool yet, that may not be feasible for you. > > - ---------------- > Robert LeBlanc > PGP Fingerprint 79A2 9CA4 6CC4 45DD A904 C70E E654 3BB2 FA62 B9F1 > > On Tue, Sep 1, 2015 at 6:19 AM, Jan Schermer wrote: > Hi, > we're in the process of changing 480G drives for 1200G drives, which should > cut the number of OSDs I have roughly to 1/3. > > My largest "volumes" pool for OpenStack volumes has 16384 PGs at the moment > and I have 36K PGs in total. That equals to ~180 PGs/OSD and would become > ~500 PG/s OSD. > > I know I can't actually decrease the number of PGs in a pool, and I'm > wondering if it's worth working around to decrease the numbers? It is > possible I'll be expanding the storage in the future, but probably not 3-fold. > > I think it's not worth bothering with and I'll just have to disable the "too > many PGs per OSD" warning if I upgrade. > > I already put some new drives in and the OSDs seem to work fine (though I had > to restart them after backfilling - they were spinning CPU for no apparent > reason). > > Your thoughts? > > Thanks > Jan > _______________________________________________ > ceph-users mailing list > [email protected] <mailto:[email protected]> > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > <http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com> > > -----BEGIN PGP SIGNATURE----- > Version: Mailvelope v1.0.2 > Comment: https://www.mailvelope.com <https://www.mailvelope.com/> > > wsFcBAEBCAAQBQJV5c1oCRDmVDuy+mK58QAAkXkP/1Wi4vBQ9BmZ6y11Eg+2 > MxFl4ajDBYosJZz1jbnRvIKWWPlVbFHxbE0cFby6RtumT6DzpRNny+12TMcE > aakwUuVR5RADh+oXzr4MU4xlPj6DWMAzSx8Bi5Mid6KVlJtK6Egsq9hCHD50 > EwXg1PcEoagJL5QOHFcT/u89TlE26Enp2cl4tjwp3ltMWj1hay+J63gpTglS > Tfmhi8hx22Q3RCWhVCFS+gWzWXjYPVfh3bONaSmK9BhqGjy98QJa6II+a6kL > gAWG7XTJl1zAKko44cj7JSqHLmzyuBfoa/PuZMOjkEfDAOW6jdTU4VUAj3bd > OK6E8sw8EMhbhlVOle6HvG1dO6bJhIt9uRxSVf+hZfFp87DoIHRAZ1J3b0PR > zB6s8b+XfSph3gnU2ZsCc3wHuqM3MFXUcI7Vn7tvdV7HWXWBTGtPhokI5COk > vgpLO1gvTTRzkNxmsLqwCTBFhFqK2zPw6xHpL1D5BcUYr/zS02+48ARZoUh7 > pRteDdsnHOPSc5m1DcldvQtQelSMgfIyULVSXlZAukIWH9rsNt7Zishj3lvR > W7z8/Ixr22TJ15mkVAAVwtlI813X59tPhmZrFmffP/GaF9vQpKUysEVZFhm1 > rrTfBt6ZBa5nhYCatojpv91HM7WNeY0XJSrl+LnwGjP9avt/B2r1SoRG61Y0 > d3BM > =J7n1 > -----END PGP SIGNATURE-----
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
