Sorry to jump into the converstation, how slow can the deletion of files actually be? One of the tests I ran a few weeks ago had me generating files, deleting them and then writing them again from a number of clients. I noticed that the space would never freed up again. I have my OSD's and their journals on dedicated partions. I had planned on asking more on this once I had a stable system again. On Mon, Mar 21, 2011 at 3:17 PM, Gregory Farnum <gregory.farnum@xxxxxxxxxxxxx> wrote: > On Sat, Mar 19, 2011 at 11:43 PM, Martin Wilderoth > <martin.wilderoth@xxxxxxxxxx> wrote: >> I have a small ceph cluster with 4 osd ( 2 disks on 2 hosts). >> >> I have been adding and removing files from the file system, mounted as ceph on an other host. >> >> Now I have removed most of the data on the file system, so I only have 300 MB left plus two snapshots. >> >> The problem is that looking at the disks the are allocating 88G of data >> on the ceph filesystem. > There are a few possibilities: > 1) You've hosted your OSDs on a partition that's shared with the rest > of the computer. In that case the reported used space will include > whatever else is on the partition, not just the Ceph files. (This can > include Ceph debug logs, so even if nothing used to be there but you > were logging on that partition that can build up pretty quickly.) > 2) You deleted the files quickly and just haven't given enough time > for the file deletion to propagate to the OSDs. Because the POSIX > filesystem is layered over an object store, this can take some time. > 3) Your snapshots contain a lot of files, so nothing (or very little) > actually got deleted. Snapshots are pretty cool but they aren't > miraculous disk space! > Given the uneven distribution of disk space I suspect option #2, but I > could be mistaken. :) Let us know! > -Greg > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
