Re: ZFS Compression

In preparation for the arrival of my HBA, I’m creating a backup of my server.   As things currently stand:

Filesystem      Size  Used Avail Use% Mounted on 
udev             79G     0   79G   0% /dev 
tmpfs            16G   86M   16G   1% /run 
rpool/root      6.9T  1.4T  5.5T  21% / 
tmpfs            79G   28K   79G   1% /dev/shm 
tmpfs           5.0M  8.0K  5.0M   1% /run/lock 
tmpfs            79G     0   79G   0% /sys/fs/cgroup 
cgmfs           100K     0  100K   0% /run/cgmanager/fs 
tmpfs            16G     0   16G   0% /run/user/1000

I’m using 1.4T.  Thats less than a formatted 2T hard drive.  That’ll definitely fit!

Except it doesn’t.  Left the rsync running overnight, got to work today, and the drive was full at approximately 1.8T.

Why?  Because apparently ZFS compression is doing it’s job..

That was a question I had regarding disk usage measurement with ZFS compression enabled.  du output is (surprise) how much space is used on the disk, not how much data you actually have.  In my case:

root@tnewman0:~# zfs get all rpool | grep compressratio                    
rpool  compressratio         1.17x                  - 
rpool  refcompressratio      1.00x                  -

1.17 x 1498796032 kilobytes is 1753591357 kilobytes, or 1.8T.  Tight fit.  Probably could have done a bit of slimming down and squeezed it in, but wheres the fun in that.

My solution:

root@tnewman0:~# zpool status 
  pool: backup 
 state: ONLINE 
  scan: none requested 
        NAME                            STATE     READ WRITE CKSUM 
        backup                          ONLINE       0     0     0 
          wwn-0x5000cca22de70c5e-part1  ONLINE       0     0     0 
errors: No known data errors 
  pool: rpool 
 state: ONLINE 
  scan: scrub repaired 0 in 2h15m with 0 errors on Mon Jan  9 22:03:44 2017 
        NAME                              STATE     READ WRITE CKSUM 
        rpool                             ONLINE       0     0     0 
          mirror-0                        ONLINE       0     0     0 
            scsi-22f6baa1200d00000-part1  ONLINE       0     0     0 
            scsi-22f4b9a2e00d00000-part1  ONLINE       0     0     0 
          mirror-1                        ONLINE       0     0     0 
            scsi-22f4be2f000d00000-part1  ONLINE       0     0     0 
            scsi-22f5b32bc00d00000-part1  ONLINE       0     0     0 
          mirror-2                        ONLINE       0     0     0 
            scsi-22f5b92a900d00000-part1  ONLINE       0     0     0 
            scsi-22f5bc2a900d00000-part1  ONLINE       0     0     0 
          mirror-3                        ONLINE       0     0     0 
            scsi-22f6b1ee800d00000-part1  ONLINE       0     0     0 
            scsi-22f6b5eb900d00000-part1  ONLINE       0     0     0 
          mirror-4                        ONLINE       0     0     0 
            scsi-22f7b0a1900d00000        ONLINE       0     0     0 
            scsi-22f7b4a0d00d00000        ONLINE       0     0     0 
          scsi-22f7bda1b00d00000          ONLINE       0     0     0 
          scsi-22f4b4ac400d00000-part1    AVAIL    
errors: No known data errors

Make a compression enabled pool on the external!

Aaaand now we wait for rsync to do its business..

UPDATE:  Interesting change in I/O wait time between filesystems.  When going from ZFS pool to EXT4, the average I/O wait percentage is ~13.14%.  When going from ZFS pool to ZFS pool, the I/O wait percentage is ~6.58%.


Leave a Reply

Your email address will not be published. Required fields are marked *