It would be lovely if there was such a RAID-filesystem... If anyone have heard of such - share with us...
eT.
--- Aaron Sherman ajs@ajs.com wrote:
We've accidentally added our spare disk to a volume. Our OEM says that this is irrevokable, and that we'll have to copy the volume to another disk tray, rebuild the volume and copy back in order to fix this. Is that true? Is there no way to shrink a filesystem and remove a disk once it's been added?
It seem contrary to the ultra-flexible spirit of the Filers....
-- Aaron Sherman Systems Architect HighWired.com -- The global high school community http://www.highwired.com
300 North Beacon Street Watertown, MA 02472 (617) 926-1850 x238 (617) 926-1861 fax asherman@highwired-inc.com / ICQ#43677395 / Y!ajstrader
"We had some good machines, but they don't work no more." -"Faded Flowers" / Shriekback
===== Yours, Eyal Traitel eTraitel@yahoo.com, Home: 972-3-5290415 (Tel Aviv) *** eTraitel - it's the new eBuzzword ! ***
__________________________________________________ Do You Yahoo!? Yahoo! Mail � Free email you can access from anywhere! http://mail.yahoo.com/
On Mon, Aug 21, 2000 at 12:38:40PM -0700, Eyal Traitel wrote:
It would be lovely if there was such a RAID-filesystem... If anyone have heard of such - share with us...
As noted by another, Veritas is capable of this. I've included an excerpt from the docs, below:
[begin vxassist man-page] NAME vxassist - create, mirror, backup, grow, shrink, and move volumes [...] shrinkto and shrinkby Decrease the length of the named volume to the length specified by newlength (shrinkto), or by the length specified by lengthchange (shrinkby). The new length or change in length is specified as a standard Volume Manager length [see vxintro(1M)]. The shrinkto opera- tion fails if the new length is not less than the current volume length. [...end]
Of course, this only shrinks the whole volume, but it is my understanding that at that point, you can re-arange the plexes so that the disk you want to remove no longer has any plexes on it. It's ugly and complex, but possible. It strikes me as very strange that a system that's otherwise so flexible does not allow this. We've really hosed ourselves because on one Filer (while attempting to set up a volume that we could volcopy to), we accidentally added disks to the root volume. So, now we have to do what? Re-install?
Of course, this only shrinks the whole volume, but it is my understanding that at that point, you can re-arange the plexes so that the disk you want to remove no longer has any plexes on it. It's ugly and complex, but possible. It strikes me as very strange that a system that's otherwise so flexible does not allow this. We've really hosed ourselves because on one Filer (while attempting to set up a volume that we could volcopy to), we accidentally added disks to the root volume. So, now we have to do what? Re-install?
-- Aaron Sherman
I had to do something similar in a case where we wanted to remove some data disks from the root volume.
You create a new volume and then copy the contents of the root volume to the new volume.
Then you use the vol options root command to make the new volume the root.
Reboot the filer. This makes the new root volume take over.
Rename the old root volume to a temp name.
Rename the new root volume to be the old root volume name.
Offline the old root volume.
Reboot the filer. This allows you to destroy the old root volume.
Use vol destroy to get rid of the old root volume.
Aaron Sherman ajs@ajs.com writes:
We've accidentally added our spare disk to a volume. Our OEM says that this is irrevokable, and that we'll have to copy the volume to another disk tray, rebuild the volume and copy back in order to fix this. Is that true? Is there no way to shrink a filesystem and remove a disk once it's been added?
It seem contrary to the ultra-flexible spirit of the Filers....
etraitel@yahoo.com (Eyal Traitel) writes:
It would be lovely if there was such a RAID-filesystem... If anyone have heard of such - share with us...
Well, there surely isn't any real difficulty in principle. It would be rather easier to implement for NetApp's RAID 4 than for some other RAID systems (easy to add a disc => easy to remove it again ...). It would go something like this:
1. A command to start "draining" one or more data discs in a volume. Filing system blocks would not get allocated on them after this. The logical size of the filing system immediately goes down to a value based on the non-draining discs.
2. Proactively move blocks allocated on the draining discs to new locations: this would be done by a sort of dummy "modification" of them. Again this would be easier with WAFL than with many other filing systems.
3. When there are no allocated blocks on the draining discs, set their contents to zero, correcting the parity disc as one goes along. One could start on this before all blocks were deallocated, of course.
4. Now, just shrink the RAID group and return the drained and zeroed discs to the spare pool.
Chris Thompson University of Cambridge Computing Service, Email: cet1@ucs.cam.ac.uk New Museums Site, Cambridge CB2 3QG, Phone: +44 1223 334715 United Kingdom.
-
Aaron Sherman -
Chris Thompson -
Eyal Traitel -
Tom Haynes