On 05 July, 2010 - Roy Sigurd Karlsbakk sent me these 1,9K bytes: > ----- Original Message ----- > > If you have one zpool consisting of only one large raidz2, then you > > have a slow raid. To reach high speed, you need maximum 8 drives in > > each raidz2. So one of the reasons it takes time, is because you have > > too many drives in your raidz2. Everything would be much faster if you > > split your zpool into two raidz2, each consisting of 7 or 8 drives. > > Then it would be fast. > > Keeping the VDEVs small is one thing, but this is about resilvering spending > far more time than reported. The same applies to scrubbing at times. > > Would it be hard to rewrite the reporting mechanisms in ZFS to report > something more likely, than just a first guess? ZFS scrub reports > tremendous times at start, but slows down after it's worked it's way > through the metadata. What ZFS is doing when the system still scrubs > after 100 hours at 100% is beyond my knowledge.
I believe it's something like this: * When starting, it notes the number of blocks to visit * .. visiting blocks ... * .. adding more data (which then will be beyond the original 100%) .. and visiting blocks ... * .. reaching the initial "last block", which since then has gotten lots of new friends afterwards. http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6899970 /Tomas -- Tomas Ögren, st...@acc.umu.se, http://www.acc.umu.se/~stric/ |- Student at Computing Science, University of Umeå `- Sysadmin at {cs,acc}.umu.se _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss