On Fri, March 19, 2010 02:28, homerun wrote: > Greetings > > I would like to get your recommendation how setup new pool. > > I have 4 new 1.5TB disks reserved to new zpool. > I planned to crow/replace existing small 4 disks ( raidz ) setup with new > bigger one. > > As new pool will be bigger and will have more personally important data to > be stored long time, i like to ask your recommendations should i create > recreate pool or just replace existing devices.
Replacing existing drives runs risks to the data -- you're deliberately reducing yourself to no redundancy for a while (while the resilver happens). It would probably be faster, and definitely safer, to back up the data, recreate the pool, and restore the data. > I have noted there is now raidz2 and been thinking witch woul be better. > A pool with 2 mirrors or one pool with 4 disks raidz2 A pool with 2 mirrors will have the same available space as a 4-disk raidz2. It will generally perform better. For small numbers of disks, I'm a big fan of using mirrors rather than RAIDZ. I've got an 8-disk hot-swap bay currently occupied by 3 2-disk pairs (with 2 slots for future expansion; maybe a hot spare, and a space to attach an additional disk during upgrades). When expanding a vdev by replacing devices, it can be done much more safely with a mirror than a RAIDZ group. With a mirror, you can attach a THIRD disk (in fact you can attach any number; one guy wrote about creating a 47-way mirror). So, instead of replacing one disk with a bigger one (eliminating your redundancy during the resilver), attach the bigger one as a third disk. When that resilver is done, you can attach the other new disk, if you have bay space; or detach one of the small disks and THEN attach the other new disk. When the second resilver is done, detach the last small disk, and you have now increased your mirror vdev size without ever reducing your redundancy below 2 copies. There's no equivalent process for a RAIDZ group. > So at least could some explain these new raidz configurations RAIDZ is "single parity" -- one drive is redundant data. A RAIDZ vdev will withstand the failure of one drive without loss of data, but NOT the failure of 2 or more. A RAIDZ pool of N drives (all the same size) has N-1 drives worth of available capacity. RAIDZ2 is "double parity" -- two drives are given to redundant data. A RAIDZ2 vdev will withstand the failure of one or two drives without loss of data, but NOT the failure of 3 or more. A RAIDZ2 pool of N drives (all the same size) has N-2 drives worth of available capacity. A problem with modern large drives is that they take a long time to "resilver" in case of failure and replacement. During that period, if you started with one redundant drive, you're down to no redundant drives, meaning that a failure during the resilver could lose your data. (This is one of the many reasons you should have backups *in addition* to using redundant vdevs). This has driven people to develop higher levels of redundancy in parity schemes, such as RAIDZ2 (and RAIDZ3). -- David Dyer-Bennet, d...@dd-b.net; http://dd-b.net/ Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/ Photos: http://dd-b.net/photography/gallery/ Dragaera: http://dragaera.info _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
