Hello guys,
I have got a storage machine with the following configuration: 2 x Intel Xeon E5-2620 CPU configuration, 64 GB of RAM, root file system on a mirror running FreeBSD 9.1 and a ZFS pool made up of 3 x 7 drives raidz2 vdevs with one SSD for L2ARC and 2 mirrored SSDs used for ZIL.
The storage pool is exported by NFS on Hypervisors that run VMs with qcow2 storage files. Every day there are certain periods when the storage experiences high load. Using gstat I can see the disks of a specific raidz vdev get 100% busy for 2-3 seconds, then they recover. This happens for 15-20 minutes and then it recovers, the load being balanced on all the disks. This generates a pretty high latency for the NFS mounts and all the VM experience load spikes.
Have you experienced these kind of issues before?
Please let me know if I can provide you more info in order to get some thoughts about what's going on.
Thanks,
Marius
I have got a storage machine with the following configuration: 2 x Intel Xeon E5-2620 CPU configuration, 64 GB of RAM, root file system on a mirror running FreeBSD 9.1 and a ZFS pool made up of 3 x 7 drives raidz2 vdevs with one SSD for L2ARC and 2 mirrored SSDs used for ZIL.
The storage pool is exported by NFS on Hypervisors that run VMs with qcow2 storage files. Every day there are certain periods when the storage experiences high load. Using gstat I can see the disks of a specific raidz vdev get 100% busy for 2-3 seconds, then they recover. This happens for 15-20 minutes and then it recovers, the load being balanced on all the disks. This generates a pretty high latency for the NFS mounts and all the VM experience load spikes.
Have you experienced these kind of issues before?
Please let me know if I can provide you more info in order to get some thoughts about what's going on.
Thanks,
Marius