No. From what I've seen, ZFS will periodically flush writes from the ZIL to disk. You may run into a "read starvation" situation where ZFS is so busy flushing to disk that you won't get reads. If you have VMs where developers expect low latency interactivity, they get unhappy. Trust me. :)
One way to address this is either have an ARC that's large enough, or add a cache-device for the zpool.
I have a config where ~20 ESX VMs share a single OpenSolaris NFS server. It has an Intel X25E for ZIL and X25M for cache. It seems to be doing ok. There are actually two of these setups. For one of them, the cache SSD died recently, and you can feel it when ZFS goes to disk for some uncached piece of data. I'll be replacing the cache SSD next week.
-Paul On 8/27/10 1:22 PM, John wrote:
Wouldn't it be possible to saturate the SSD ZIL with enough backlogged sync writes? What I mean is, doesn't the ZIL eventually need to make it to the pool, and if the pool as a whole (spinning disks) can't keep up with 30+ vm's of write requests, couldn't you fill up the ZIL that way?
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
