On Mon, 8 Sep 2014 00:20:37 +0000 (UTC) Alex Moore wrote: > Andrei Mikhailovsky <andrei@...> writes: > > > > > Hello guys, > > > > was wondering if it is a good idea to enable TRIM (mount option > > discard) > on the ssd disks which are used for > > either cache pool or osd journals? > > As far as the journals are concerned, isn't this irrelevant if you're > assigning a block device to them (which I believe is recommended)? As > there's no associated filesystem with which to use the "discard" mount > option, nor call "fstrim". > Precisely. And even if the journal were to be on a filesystem (which indeed isn't recommended), it would be one big blob, once allocated and never freed, thus rendering trim pointless.
> > > > For performance, is it better to enable it or run fstrim with cron > > every > once in a while? > > My belief is that for the journals you want to "under provision" the SSD > - eg perhaps by creating a partition that is a fair bit smaller than the > SSD itself, and using that as the journal. > Indeed. This will both allow the SSD to find plenty of free cells to rotate through and also prolong its life. Of course if you're using Intel DC 3700S SSDs you probably won't get any speed benefit from this at least, they're never slowing down. ^^ > And for the OSD (whether or not it's part of a cache tier), which does > have a filesystem, I noticed a significant performance deterioration > when using the "discard" mount option versus occasionally running > "fstrim". A sentiment often heard and one I can confirm as well. >I have tried both BTRFS and XFS, and it was similar in both > cases. That said, my understanding is that SATA 3.1 has some > optimizations (queued TRIM) that should help reduce the negative effects > of mounting with the "discard" option, but I can't confirm from > experience as the chipsets on my systems don't support it. > Wasn't aware of that yet, thanks for that information. But yeah, no support for that in any of my systems either and probably not for some time either. ^^ Christian > > > > Thanks > > > > Andrei > > > > > > > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > -- Christian Balzer Network/Systems Engineer ch...@gol.com Global OnLine Japan/Fusion Communications http://www.gol.com/ _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
