I've got a server at home I use to store personal media and stream it to my TV with Plex. I was having issues with lagging during streaming. After some inspection, it didn't seem to be a constraint on CPU resources causing the problem, but IO. After disabling other IO tasks on the server, performance didn't improve very much. I thought I might be having an upcoming disk failure, but the smart data on all of the drives was good and hdparm gave me about what I expected the raw read speed to be. On a whim, I checked the filesystem fragmentation with xfs_db and found it was 98% fragmented. Holy cow!

~> sudo xfs_db -r /dev/mapper/vg1-raid
xfs_db> frag
actual 2645859, ideal 32191, fragmentation factor 98.78%
Note, this number is largely meaningless.
Files on this filesystem average 82.19 extents per file

I ran xfs_fsr based on recommendations I found online and waited.

~> sudo xfs_fsr
xfs_fsr -m /proc/mounts -t 7200 -f /var/tmp/.fsrlast_xfs ...
/ start inode=0
/usr start inode=0
/tmp start inode=0
/var start inode=0
/raid start inode=0
xfs_fsr startpass 0, endpass 0, time 7306 seconds

Roughly two hours later, fragmentation had only decreased slightly.

~> sudo xfs_db -r /dev/mapper/vg1-raid
xfs_db> frag
actual 2343029, ideal 32191, fragmentation factor 98.63%
Note, this number is largely meaningless.
Files on this filesystem average 72.79 extents per file

Reading the manpage for xfs_fsr I found that the default invocation defrags all xfs filesystems for a maximum of two hours. When you run it with a device, it will run ten passes by default. I next ran xfs_fsr /dev/mapper/vg1-raid and waited a very long time—roughly twelve hours. When xfs_fsr completed, the drive was down to 1.03% fragmentation. Sweet!

~> sudo xfs_db -r /dev/mapper/vg1-raid
xfs_db> frag
actual 32526, ideal 32192, fragmentation factor 1.03%
Note, this number is largely meaningless.
Files on this filesystem average 1.01 extents per file

Three Months Later

Three months later, the filesystem is quite fragmented again!

~> sudo xfs_db -r /dev/mapper/vg1-raid
xfs_db> frag
actual 197628, ideal 26411, fragmentation factor 86.64%
Note, this number is largely meaningless.
Files on this filesystem average 7.48 extents per file

But why was it so fragmented?

I've long heard praises of how great Linux filesystems (such as ext4 and xfs) are at avoiding fragmentation. I can't think of how this filesystem could have got so fragmented. If you have any hypotheses, I'd really appreciate an email!