On Tue, Jun 16, 2015, at 10:29 PM, Tecnologia Charne.Net wrote: > Problem was btrfs metadata: > > cyrus squatter user/jav...@unnoba.edu.ar > Writing index update: No space left on device > > > Filesystem checks: > > mda:~# btrfs filesystem df /var/spool/cyrus/ > Data, single: total=1.26TiB, used=349.31GiB > System, DUP: total=8.00MiB, used=160.00KiB > System, single: total=4.00MiB, used=0.00B > *Metadata, DUP: total=3.50GiB, used=2.99GiB* > Metadata, single: total=8.00MiB, used=0.00B > > > btrfs filesystem usage /var/spool/cyrus > Overall: > Device size: 1.27TiB > Device allocated: 1.27TiB > Device unallocated: 60.00KiB > Device missing: 0.00B > Used: 355.30GiB > Free (estimated): 943.69GiB (min: 943.69GiB) > Data ratio: 1.00 > Metadata ratio: 2.00 > Global reserve: 0.00B (used: 0.00B) > > Data,single: Size:1.26TiB, Used:349.31GiB > /dev/xvdb1 1.26TiB > > Metadata,single: Size:8.00MiB, Used:0.00B > /dev/xvdb1 8.00MiB > > *Metadata,DUP: Size:3.50GiB, Used:2.99GiB** ** /dev/xvdb1 7.00GiB* > > System,single: Size:4.00MiB, Used:0.00B > /dev/xvdb1 4.00MiB > > System,DUP: Size:8.00MiB, Used:160.00KiB > /dev/xvdb1 16.00MiB > > Unallocated: > /dev/xvdb1 60.00KiB > > > # btrfs balance start -v -dusage=5 /var/spool/cyrus/ > Dumping filters: flags 0x1, state 0x0, force is off > DATA (flags 0x2): balancing, usage=5 > Done, had to relocate 213 out of 1302 chunks > > > Now, it's working.
Why would anybody ever run production data on btrfs? https://btrfs.wiki.kernel.org/index.php/Main_Page "The Btrfs code base is under heavy development." https://btrfs.wiki.kernel.org/index.php/Gotchas "Files with a lot of random writes can become heavily fragmented (10000+ extents) causing trashing on HDDs and excessive multi-second spikes of CPU load on systems with an SSD or large amount a RAM." Unfortunately, both cyrus.index files and skiplist/twoskip databases get a lot of random writes - though both repack occasionally which reduces fragmentation. I've tried other filesystems a few times, and I keep coming back to ext4 for consistent performance and crash safety (at least on decent hardware which doesn't lie about fsync) - XFS has lost data, btrfs has always been shiny at first, but turned into a performance disaster over time. -- Bron Gondwana br...@fastmail.fm