Migrating to ZFS
in [Solaris]
|
Prev: Help needed because C++ free() does not give the memory back to the system
Next: format probelm - Warning: Non-contiguous partition (0) in table.
From: David Combs on 29 Oct 2009 00:55 In article <4ae1ac77$0$83234$e4fe514c(a)news.xs4all.nl>, Casper H.S. Dik <Casper.Dik(a)Sun.COM> wrote: >chuckers <chuckersjp(a)gmail.com> writes: > >>So that goes back to an earlier question. Does it make sense to >>remove SDS from the disks and repartition >>c0t1d0s0 so that it is the whole disk and use that? That would make >>the root pool use the entire 132GB. >>Is that a good idea? How do I get my data off of c0t0d0s5 and >>c0t0d0s6 onto the root pool file system? > >Possible, but perhaps you want the data in a different pool. > >I'm still using actual swap partitions/slices and I'm not using ZFS >for swap. > >>Is it better to have a seperate root pool and data pool? Or is >>tossing everything onto the root pool sufficient? >>Is that dependent on whether I have only 2 disks or lots and lots of >>disks? > >I would use a rpool and different pools for data because an rpool can only >be a simple device or simple mirror. Casper, a question: What's the REASON for that rule, that restriction, for root-pools only? Thanks! David
From: David Combs on 29 Oct 2009 01:10 In article <4ae1801f$0$83232$e4fe514c(a)news.xs4all.nl>, Casper H.S. Dik <Casper.Dik(a)Sun.COM> wrote: >chuckers <chuckersjp(a)gmail.com> writes: > >>What has me confused is the fact that the root pool for ZFS has to be >>done on a slice rather than the entire >>disk (as stated on p. 127 of the manual above.) > >The reason is that it's not possible to boot from EFI labeled disk; >you need to use a standard SMI labeled disk. > >>The manual recommends using the entire disk for ZFS pools but root >>files system pools have to be based on >>slices (very confusing.) When converting to a ZFS root files sytem, >>does it format the disk such that the >>slice becomes the entire disk? Or do I have to do that myself? > >If you give a whole disk to ZFS, it will "label" the disk with a EFI >label (wrong in your case). QUESTION: Why is that wrong in his case? THANKS! > >If you use a "slice", ZFS will not change the size of the slice nor >will it change the restof the disk. > >>Does the remaining 128 GB go to waste? Can I take the remaing 128 GB >>and make it into a seperate >>ZFS disk pool by changing the remaining disk space into one >>partition? Should I do that? > >I think 8GB is on the "light" case; specifically, if you want to upgrade >using live upgrade, then you want to have sufficient space to >clone and upgrade. You can use compression (only using the default >algorithm) > >>c0t1d0s0 doesn't start at cylinder 1. Swap does. If I use s0, and >>ZFS doesn't expand the slice for me, >>I have non-contiguous cylinders for the remaining 128 GB. If I want >>to make a disk pool out of that space, >>can I do it based on two non-contiguous slices? Is that a wise thing >>to do? > >Not possible (well, it is possible but not for a root pool). > >A root pool needs to be: > > - in a slice, not a full disk QUESTION: WHY? > - a "simple" vdev or a mirror (two or more vdevs); not a > concatination > - compression is allowed but only using the default algorith, > >>If ZFS expands the slice to the entire disk, what happens to the my >>mirror halfs on d5 and d6? > >It won't do that. > >>Should I remove SDS first and then attempt a Live upgrade after >>repartitioning c0t1d0s0 to be the entire disk? 1: What's "SDS"? (inane question, yes, so THANK YOU) 2: That "best practices" zfs-document, I think it said something about deliberately NOT using the entire disk -- use a little less. The reason was that if you wanted it in a mirror, or if it crashed, you'd have to buy another EXACTLY THE SAME -- same manufacturer, same model, even same software upgrade. (Not that I understand WHY the restriction, given just how fancy, clever, ingenious, etc zfs seems to be.) The reason for not using 100% of a disk is that maybe the replacement disk had some bad blocks -- eg MORE bad blocks than the original, or maybe the other side of the mirror, I don't recall. Anyway, the new disk would then look (be) *smaller* than the other, ie NOT identical. Something like that, I recall. Now, if anyone can flesh that out, YOU can -- and much THANKS for that! David > >I don't think you can "remove" SDS without losing the data. >(I don't think you can downgrade SDS to a simple UFS filesystem) > > >Casper >-- >Expressed in this posting are my opinions. They are in no way related >to opinions held by my employer, Sun Microsystems. >Statements on Sun products included here are not gospel and may >be fiction rather than truth.
From: David Combs on 29 Oct 2009 01:21 In article <92e2b766-c3a3-41e7-ac98-f7def791b884(a)d23g2000vbm.googlegroups.com>, ITguy <southallc(a)gmail.com> wrote: >Here's what I'd do: > >1) For each SVM mirror, detach the submirror from c0t1d0 >2) Format c0t1d0 using 100% of the disk as slice 0. >3) Create the root ZFS pool using c0t1d0s0 >4) Live upgrade to the ZFS pool, using a Zvol in the root pool for >swap >5) Create ZFS datasets in the root pool for /data and /data1 >6) Transfer files from /data and /data1 to the new ZFS datasets >7) Activate the new BE and boot your new ZFS root >8) If the new BE looks good, re-label c0t0d0 to match c0t1d0 >9) Add c0t0d0s0 as a mirror for the root pool. > 1: Do you show any of this in vfstab? 2: If not, then is there some standard file, in /etc/, say, where you *document* this stuff. 3: Ditto for the pools and zfs's and dir-trees also containing, mounted here and there, other zfs's? (Would such a tree consist of ONLY zfs's, ie would it be best to NOT files and zfs's populating the same tree (well, directly owned by the same "top level" zfs, ie pool/zfs. Or would want to have a zfs contain only EITHER files OR zfs's. Like, what kinds of designs do (good) sysadmins typically try to end up with? THANKS! David
From: Ian Collins on 29 Oct 2009 02:16 David Combs wrote: > In article <4ae1acf9$0$83234$e4fe514c(a)news.xs4all.nl>, > Casper H.S. Dik <Casper.Dik(a)Sun.COM> wrote: >> chuckers <chuckersjp(a)gmail.com> writes: >> >>> So I can do this: >>> zpool create rpool c0t1d0s0 >>> (will add c0t0d0s0 as a mirror after upgrade.) >>> zpool create datapool c0t1d0s1 >>> (will add c0t0d0s1 as a mirror after upgrade.) >> I have a similar setup on some of my systems. I also upgrade from >> a UFS "/" partition, before I already had a ZFS "data" pool. >> >> > .... > > Naive question: > > Does that mean that before you upgraded you had a UFS > root partition, and when you finished you had a > zfs root pool, no more ufs? > > What advantage from having both ufs and zfs on the > same computer ("system" is the current vocab, I guess?)? Reverse the question: why would you want both on the same system? -- Ian Collins
From: Ian Collins on 29 Oct 2009 02:17
David Combs wrote: > In article <4ae1ac77$0$83234$e4fe514c(a)news.xs4all.nl>, > Casper H.S. Dik <Casper.Dik(a)Sun.COM> wrote: >> chuckers <chuckersjp(a)gmail.com> writes: >> >>> So that goes back to an earlier question. Does it make sense to >>> remove SDS from the disks and repartition >>> c0t1d0s0 so that it is the whole disk and use that? That would make >>> the root pool use the entire 132GB. >>> Is that a good idea? How do I get my data off of c0t0d0s5 and >>> c0t0d0s6 onto the root pool file system? >> Possible, but perhaps you want the data in a different pool. >> >> I'm still using actual swap partitions/slices and I'm not using ZFS >> for swap. >> >>> Is it better to have a seperate root pool and data pool? Or is >>> tossing everything onto the root pool sufficient? >>> Is that dependent on whether I have only 2 disks or lots and lots of >>> disks? >> I would use a rpool and different pools for data because an rpool can only >> be a simple device or simple mirror. > > Casper, a question: What's the REASON for that rule, that restriction, > for root-pools only? Booting. -- Ian Collins |