Folks, One of my colleagues asked something about the scrub command, and I'm not quite sure as to the right answer. If I'm reading things correctly, scrub doesn't have any way of being interrupted and then continuing. That is, if I run 'zfs scrub tank', and then kill it before it completes, a subsequent run will start all over again, right?
The reason behind this is that we'd like to investigate writing a daemon that does continuous scrub, such that it's smart enough to suspend scrubbing when I/O goes above a certain threshold, and then continues again where it left off at a later time, and also simply starts again at the beginning of the pool once it reaches the end. The rationalization is that we don't really have a whole lot of time where there's very little activity on our pools, but there's a fair amount of time where pool activity is well under a moderate utilization level. Thus, it would be nice if we could always scrub where utilization was (say) 20% or less, and stop anytime else. The problem for us is exacerbated by the type of pools - huge numbers of tiny files, so it takes quite a long time to scrub several TB - I think a full, no-load scrub is on the order of 12-16 hours. -- Erik Trimble Java System Support Mailstop: usca22-123 Phone: x17195 Santa Clara, CA Timezone: US/Pacific (GMT-0800)
