James Dickens wrote:
eric was allready talking about printing the last time a disk was accessed when a disk was about to be imported, my idea would be run that check twice, once initially and if it looks like it could be still in use, like the pool wasn't exported and last write occurred in the last 30 seconds or a few minutes, it would sleep, then run the check again, it the drive showed up changed ( even if it was just because of an updated atime) the import would fail with EBUSY. The drives would all ready be spun up during all of this because were importing them.
You're assuming that the system that owns the pool is up and running at the times you're making your checks. Or that a certain span of wall clock time is a good indication of what is in use.
In any situation where two entities are unclear as to device ownership they go to a third entity. You can say thats a cluster node or that its a somewhat intelligent person. However, trying to work around the issue is going to lead to a disaster at some point. Seen it in action plenty of times. If people want to use ZFS commands to try and get around the issue then have at it but please don't ask the folks working on ZFS to assemble the gun, purchase the ammo, load the weapon, and point it in the general direction of your lower locomotive extremities. :)
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
