![]() ![]() If you also have to play with partition schemes this gets much more involved and further raises the bar as to the level of the admin supporting the subsystem. When all members are replaced you will have double the capacity available for use. it allocates sectors as needed) you can create a vdev based on say a couple 1TB drives, you can then remove those drives (one at a time) and replace them with say 2TB drives. Since ZFS carves up a block device into discrete atomic units (i.e.If on the other hand you don't have any partition table your lun and zfs will /ALWAYS/ be aligned with no manual/end user intervention. But then if the back-end san changes the array configuration now you have the issue of being mis-aligned again. To your point you can manually re-size this GPT offset to say 768KiB or 1536KiB. Now you effectively have a stripe width of 384KiB. ![]() say you have a san which has luns comprised of 3D+1P (raid 5 of 4 disks). However you then have the secondary issue of the stripe width alignment. For example with a 1024KiB offset that will work to fit stripe sizes of 128, 256, 512, or 1024KiB (since ZFS itself uses 128KiB sizes that would be the smallest). with any GPT offset there is a problem in proper alignment both to the stripe size of an underlaying device as well as it's stripe width.Besides the bug that is secondary to the discussion (but needs to be fixed, the 1049KiB offset opposed to 1024KiB), there are two issues. ![]()
0 Comments
Leave a Reply. |