I have just upgraded my jumpstart server to S10 u2 b9a. It is an Ultra 10 with two 120GB EIDE drives. The second drive (disk1) is new, and has u2b9a installed on a slice, with most of the space in slice 7 for the ZFS pool I created pool1 on disk1, and created the filesystem pool1/ro (for legacy reasons). I them moved my data from the original disk0 UFS file system to pool1/ro. Initially I thought to keep the directory layout the same but now I realize that ZFS gives me more flexibility so I am mv''ing things around. I created as second filesystem pool1/jumpstart, and deviced to mv pool1/ro/jumpstart/* to pool1/jumpstart. All the data is staying in the same pool. No data is actually getting changed, it is just being relocated. If this were a UFS filesystem, the move would be done before I even finish writing this. However, with ZFS, this is taking a long time. Is there a more efficient way of doing this? It sure looks like all 70GB are getting copied. Thanks Steffen
Steffen Weiberle wrote:> I created as second filesystem pool1/jumpstart, and deviced to mv > pool1/ro/jumpstart/* to pool1/jumpstart. All the data is staying in the > same pool. No data is actually getting changed, it is just being > relocated. If this were a UFS filesystem, the move would be done before > I even finish writing this. However, with ZFS, this is taking a long time.It may well be in the same pool but it is in a different file system so it has to copy and delete to move it. This is covered by 6400399 (http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6400399) -- Darren J Moffat
Darren J Moffat wrote On 06/12/06 09:09,:> Steffen Weiberle wrote: > >> I created as second filesystem pool1/jumpstart, and deviced to mv >> pool1/ro/jumpstart/* to pool1/jumpstart. All the data is staying in >> the same pool. No data is actually getting changed, it is just being >> relocated. If this were a UFS filesystem, the move would be done >> before I even finish writing this. However, with ZFS, this is taking a >> long time. > > > It may well be in the same pool but it is in a different file system so > it has to copy and delete to move it. > > This is covered by 6400399 > (http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6400399) >Thanks, Darren. Guess I''ll have to wait the 4+ hours for the data to be moved :( I did the original move overnight and did not measure how long that took. But since it was from one spindle to another, I imagine it did better than within the same spindle! Thanks!