Strange performance characteristics with ZFS
Steven Schlansker
stevenschlansker at berkeley.edu
Mon Jul 16 19:48:58 UTC 2007
Hello everyone,
I'm experiencing what (to me) seems to be strange performance with
ZFS. I have it set up as follows:
[steven at universe /universe]$ sudo zpool status
pool: universe
state: ONLINE
scrub: resilver completed with 0 errors on Mon Jul 16 10:33:52 2007
config:
NAME STATE READ WRITE CKSUM
universe ONLINE 0 0 0
raidz1 ONLINE 0 0 0
ad10 ONLINE 0 0 0
ad6 ONLINE 0 0 0
ad4 ONLINE 0 0 0
raidz1 ONLINE 0 0 0
ad12 ONLINE 0 0 0
ad16 ONLINE 0 0 0
ad14 ONLINE 0 0 0
I share it over NFS to a linux machine. I'm trying to write a very
large (30GB) disk image to it over nfs, and it goes in short fits and
bursts.
It copies a few (mega?)bytes of data (not sure how much exactly), and
top shows:
605 root 1 4 0 3616K 1068K - 1:17 5.18% nfsd
603 root 1 4 0 3616K 1068K - 1:33 4.69% nfsd
602 root 1 4 0 3616K 1068K - 1:09 4.35% nfsd
604 root 1 4 0 3616K 1068K - 1:24 3.66% nfsd
Then, the copy on the linux box freezes entirely and top shows:
603 root 1 -4 0 3616K 1068K zfs:(& 1:34 4.98% nfsd
605 root 1 -4 0 3616K 1068K zfs 1:17 4.25% nfsd
604 root 1 -4 0 3616K 1068K zfs 1:24 4.05% nfsd
602 root 1 -4 0 3616K 1068K zfs 1:10 3.96% nfsd
While the nfsds are stuck in the 'zfs' state, ssh is very
unresponsive (few seconds per keystroke)
After a few seconds of being hung like this, the copy resumes and the
nfsds go back to a blank state.
I guess the ssh hanging is the most annoying symptom, as it makes it
hard to use the system while copying. It would be nice to speed the
transfers up too though.
Is this expected behavior? Is there some way I can make it run
faster/better?
[steven at universe /universe]$ uname -a
FreeBSD universe.stevenschlansker.is-a-geek.org 7.0-CURRENT FreeBSD
7.0-CURRENT #1: Thu Jul 12 11:48:11 PDT 2007
root at universe.stevenschlansker.is-a-geek.org:/usr/obj/usr/src/sys/
GENERIC amd64
Thanks,
Steven
More information about the freebsd-current
mailing list