[zfs-discuss] zpool iostat bandwidth bug?

Dan Swartzendruber dswartz at druber.com
Tue Nov 27 11:12:49 EST 2012


It seems almost like a write that is 'too compressed' isn't even 
counted.  Did a test on a compress=on dataset.

1. Copy an 8GB file from /dev/zero.  Lightning fast (as expected).

root at proxmox:/foo/bar# zfs get compressratio foo/bar
NAME     PROPERTY       VALUE  SOURCE
foo/bar  compressratio  1.00x  -

2. Copy over /usr from the root ext4 filesystem

root at proxmox:/foo/bar# zfs get compressratio foo/bar
NAME     PROPERTY       VALUE  SOURCE
foo/bar  compressratio  1.74x  -

3. Remove /tank/foo/bar/usr

root at proxmox:/foo/bar# zfs get compressratio foo/bar
NAME     PROPERTY       VALUE  SOURCE
foo/bar  compressratio  1.45x  -

If I delete the 8GB all-zeroes file instead of the usr directory, the 
compressratio doesn't change (e.g. stays at 1.74).



More information about the zfs-discuss mailing list