OpenSolaris build 128 now availble - zfs dedup in it

I've being waiting for this update to try deduplication of zfs. Being knowing that a lot of files will be written as result of image update I set zfs compression including rpool file system (except swap and dump) before doing update. Zfs compression on my home files already being for one month and proved to save space/time/power.

Here what I have with zfs compression after all.

$ zfs get -r compressratio
NAME                              PROPERTY       VALUE  SOURCE
rpool                             compressratio  1.14x  -
rpool/ROOT                        compressratio  1.05x  -
rpool/ROOT/b127                   compressratio  1.93x  -
rpool/ROOT/b128a                  compressratio  1.05x  -
rpool/roman                       compressratio  1.18x  -

Now updating image. To get recent bits

$ pfexec pkg image-update --be-name b128a
 

Reboot in new boot environment crossing fingers ;)

To update zfs to most recent version

$ pfexec zpool upgrade rpool

Remembering old problem I did in advance. Check device name of your zfs.

$ pfexec installgrub -m /boot/grub/stage1 /boot/grub/stage2 /dev/rdsk/c4t0d0s0


First hours impression:

1. time slider daemon is broken. Bug and fix here.

2. Firefox now 3.5.5

3. zfs now have deduplication.

$ zfs get dedup rpool

I am going to set it to sha256,verify. The reason for verify is that I am paranoid about my data. I do believe that:

  • If you have files born as result of cp then it is fine to no verify. In this case you do believe that you have identical files.
  • In my case data comes from different sources. If two blocks got the same sha256 then I do believe that it is collision. So I would like to verify if they are actually the same.
$ pfexec zfs set dedup=sha256,verify rpool
$ pfexec zfs set dedup=off rpool/swap
$ pfexec zfs set dedup=off rpool/dump

Time and Math will prove who is right. Anything that can possibly go wrong, does.

Comments:

Saving 5% is nice but nothing compared to the 93% on your b127 partition.

Posted by Ramon van Belzen on December 05, 2009 at 11:25 PM PST #

Hi Ramon, I do believe that this is due to the fact that b127 was snapshotted and then cloned to became b128. Now b127 takes only 22,0M on the disk. Current status is the following. Deduplication already save ~5G on my disk. Compression saves ~11GB. Need to notice that only recently written blocks are compressed and deduped.

$ zpool list
NAME SIZE ALLOC FREE CAP DEDUP HEALTH ALTROOT
rpool 140G 66,1G 73,9G 47% 1.03x DEGRADED -

$ zfs list
NAME USED AVAIL REFER MOUNTPOINT
rpool 66,9G 71,0G 64K /rpool
rpool/ROOT 19,6G 71,0G 18K legacy
rpool/ROOT/b127 22,0M 71,0G 10,2G legacy
rpool/ROOT/b128a 19,6G 71,0G 10,4G legacy
rpool/dump 1023M 71,0G 1023M -
rpool/riXXXXXX 45,3G 71,0G 34,8G /export/home/riXXXXXX
rpool/swap 1023M 71,8G 275M -

$ zfs get -r compressratio|grep -v snap
NAME PROPERTY VALUE SOURCE
rpool compressratio 1.14x -
rpool/ROOT compressratio 1.06x -
rpool/ROOT/b127 compressratio 1.93x -
rpool/ROOT/b128a compressratio 1.06x -
rpool/ROOT/b128a@2009-09-04-11:28:13 compressratio 1.00x -
rpool/ROOT/b128a@2009-12-04-12:16:20 compressratio 1.02x -
rpool/dump compressratio 1.00x -
rpool/riXXXXXX compressratio 1.17x -
rpool/swap compressratio 1.00x -

Posted by Roman Ivanov on December 06, 2009 at 10:35 PM PST #

Post a Comment:
  • HTML Syntax: NOT allowed
About

Roman (pomah) Ivanov, ISV Engineering. Tips how to run Oracle best on Sun. Performance monitoring and tuning for system administrators. OpenSolaris user experience.

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today