Anton B. Rang wrote:
If the SCSI commands hang forever, then there is nothing that ZFS can
do, as a single write will never return. The more likely case is that
the commands are continually timining out with very long response times,
and ZFS will continue to talk to them forever.
It looks like the sd driver defaults to a 60-second timeout, which is
quite long. It might be useful if FMA saw a potential fault for any I/O
longer than some much lower value.
We did a rather extensive study of this a few years ago.
It turns out that 60 seconds is appropriate for most systems,
and we failed to justify changing it as the default. You can,
of course, tune it yourself, but you don't want to have an
instable system.
(This gets tricky with power management, since if you have to wait for
the disk to spin up, it can take a long time compared to normal I/O.)
Yes, even without power management you may need to wait for
spin up. Most disks spec around 30 seconds for spin up. But
there are some devices such as tapes and removable media for
which more than 30 seconds, but less than 60 seems reasonable.
Similarly, the number of retries is tunable. Some combination of
timeout/retry may suit a specific situation better, but that is
difficult to predict for the general case.
That said, it sounds to me like your enclosure is actually powering down
the drive. If so, it ought to stop responding to selection, and I/O should
fail in a "hard" way within 250 ms (or less, depending on whether you've
got a SCSI bus which supports QAS, as the newer, faster versions do).
Agree. For a spun down drive, an iop will cause it to spin up.
You will still be talking to the electronics, just waiting on the
media.
-- richard
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
