Andrew Hisgen
2008-Nov-17 18:29 UTC
[zfs-discuss] proper or best zpool command to use to resurrected a disk that has been temporarily disconnected
Consider the following scenario. A disk that is part of a zfs zpool and is also part of a mirror in that zpool becomes disconnected. Imagine that the disconnection last minutes or hours or days. Then imagine that the disk becomes reconnected, but, in a manner where hot-plug event does not cause it to come back online automatically. (If it helps for concreteness, let us say that the disk is implemented as an iSCSI target provided by some other host/system). What is the proper (or best if there is more than one way) zpool command to use to bring the disk back such that zfs will incrementally resilver it as necessary (using the Dirty Time Log or whatever)? The candidate commands would seem to be: zpool online But not clear if this is a candidate because seemingly it is the inverse of "offline", and offline is said to be an explicit action. zpool clear Will the act of clearing errors cause ZFS to resume using the device, and to incrementally resilver off of the Dirty Time Log? zfs replace But the text of the manpage seems to say that replace is alot like an attach followed by a detach. The manpage prose also treats the case where the disk has literally been replace, that is, by a completely different disk; the manpage does not explicitly mention the case where the disk is the exact same physical disk. So, seemingly, replace could end up resilvering the entire disk, not using the Dirty Time Log? zfs scrub Not a good candidate because it will read the entire disk (or more precisely, all blocks that are reachable). thanks, Andy -- Email: Andrew.Hisgen at sun.com
Miles Nordin
2008-Nov-17 18:52 UTC
[zfs-discuss] proper or best zpool command to use to resurrected a disk that has been temporarily disconnected
>>>>> "ah" == Andrew Hisgen <Andrew.Hisgen at Sun.COM> writes:ah> (If it helps for concreteness, let us say that the disk is ah> implemented as an iSCSI target for me, iSCSI targets do come back online automatically. then, they don''t resilver, or don''t resilver enough. If they were gone for too long, after a few hours or days they get marked FAULTED for cksum errors. then I ''zpool clear'' them, and they resilver a little more. then I ''zpool scrub'', and they finish resilvering. Unless there are too many CKSUM errors---then they get marked FAULTED again, and I have to start over. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 304 bytes Desc: not available URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20081117/f2bae794/attachment.bin>