You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Have a problem with ZFS hanging on high load ( local iozone 64 threads at same time remote NFS3 client running 128 threads ).
ZFS Server is running Ubuntu LTS 10.04
Kernel: 2.6.38-11-server #50~lucid1-Ubuntu SMP Tue Sep 13 22:10:53 UTC 2011 x86_64 GNU/Linux
ZFS is from PPA: https://launchpad.net/~zfs-native/+archive/stable - version: 0.6.0.33
The ZFS server has hyperthreading disabled, 48GB RAM, drives are JBOD's on 2 megaraid_sas controllers, the raid2z is sata 2TB, the ZIL drives on the other controller are sas.
The nfs client is a similar server, and they are connected via 10Gb NICS ( ixgbe )
Locally am running:
while true; do iozone -l 64 -o -O -s 32m -r 4k; sleep 1; done
At the same time, another server is running this to a different zfs filesystem ( same pool ) via nfs client:
while true; do iozone -l 128 -o -O -s 32m -r 4k; sleep 1; done
During testing load average sits at about 80, lots of space free on the pool.
Test runs for about an 1 hour before the local iozone causes:
Once the issue has occured, there is no further io, will respond to pings, no new ssh, nfs client hangs, existing ssh still alive but would not reboot, have to press button.
Am able to replicate the failure, happy to provide info or assist.
The text was updated successfully, but these errors were encountered:
Hi All,
Have a problem with ZFS hanging on high load ( local iozone 64 threads at same time remote NFS3 client running 128 threads ).
ZFS Server is running Ubuntu LTS 10.04
Kernel: 2.6.38-11-server #50~lucid1-Ubuntu SMP Tue Sep 13 22:10:53 UTC 2011 x86_64 GNU/Linux
ZFS is from PPA: https://launchpad.net/~zfs-native/+archive/stable - version: 0.6.0.33
The ZFS server has hyperthreading disabled, 48GB RAM, drives are JBOD's on 2 megaraid_sas controllers, the raid2z is sata 2TB, the ZIL drives on the other controller are sas.
The nfs client is a similar server, and they are connected via 10Gb NICS ( ixgbe )
Locally am running:
At the same time, another server is running this to a different zfs filesystem ( same pool ) via nfs client:
During testing load average sits at about 80, lots of space free on the pool.
Test runs for about an 1 hour before the local iozone causes:
Here is another, this once has matching arcstat below:
Some arstats at the time of failure:
The pool consists of:
Once the issue has occured, there is no further io, will respond to pings, no new ssh, nfs client hangs, existing ssh still alive but would not reboot, have to press button.
Am able to replicate the failure, happy to provide info or assist.
The text was updated successfully, but these errors were encountered: