It's a 250GB SSD using btrfs, here is the "df -h" output

Filesystem                 Size  Used Avail Use% Mounted on
/dev/sdb1                  233G  103G     0 100% /mnt/cache

Last night I was able to delete a few files on the drive and it jumped back to "40%-ish" use and but this morning I couldn't even get that to work. Is there something that causes df not to report the right usage? Only thing I can add out is I just moved my plex config folder to the cache drive and it has a billion little files in there (Trying to run du on the plex folder takes hours).

Never seen it on a cache device, but it does happen sometimes on array devices, on those cases it was fixed by copying a large file locally from another disk, e.g., use cp or midnight commander and copy a large file from any disk to the cache device (if you copy from a user share use /mnt/user0).

@johnnie.black I tried creating a dummy file (with dd) and copying a file and I got:

 

cp: cannot create regular file 'XXX': No space left on device

 

Both times.

Post your diagnostics.

@johnnie.black Here you go

tower-diagnostics-20170828-1027.zip

Crap, I just noticed the /var/log mount is full and so nothing is writing to it anymore. I cleared out space but in the past I've had to restart before logging resumes so I'm going to do that now.

Filesystem is not full but it's fully allocated, see here:

 

 

Ok, I have run:

btrfs balance start -dusage=5 /mnt/cache

And that completed with the message:

 

Done, had to relocate 38 out of 236 chunks

For the following values:

 

Data, single: total=196.87GiB, used=98.62GiB
System, single: total=4.00MiB, used=48.00KiB
Metadata, single: total=2.01GiB, used=861.39MiB
GlobalReserve, single: total=157.81MiB, used=0.00B

Which value do I need to watch to know when to re-run that?

15 minutes ago, joshstrange said:

Data, single: total=196.87GiB

 

This is the one to watch, when it's getting close to the device size you should run a balance, this is how it was:

 

Data, single: total=231.87GiB, used=101.35GiB

You can see it best on btrfs fs show:

 

Quote

Label: none  uuid: 93e1b729-262a-4f89-a3c3-1279794de80f
    Total devices 1 FS bytes used 102.20GiB
    devid    1 size 232.89GiB used 232.89GiB path /dev/sdb1

 

size is total device size, used is allocated space.

Thank you @johnnie.black!

Also, when all space is allocated the first balance should be run with -dusage=5 or it can fail, but now you can run it with an higher value to recover most allocated space, e.g.:

 

btrfs balance start -dusage=75 /mnt/cache

Edited August 28, 20178 yr by johnnie.black