Thanks to:
"Alan Rollow - Dr. File System's Home for Wayward Inodes."
Kurt Carlson <snkac_at_java.sois.alaska.edu>
Steve Gwynn <steve.gwynn_at_mci.com>
Peter Stern <peter_at_wiscpa.weizmann.ac.il>
Lucien_HERCAUD_at_paribas.com
Dejan Muhamedagic <dejan_at_yunix.co.yu>
George Guethlein <GGuethlein_at_GiantOfMaryland.com>
"Dr. Tom Blinn, 603-884-0646" <tpb_at_zk3.dec.com>
Oisin McGuinness <oisin_at_sbcm.com>
Ajay Nautiyal <sxan_at_java.sois.alaska.edu>
Original Question:
> I have a disk rz1c which is my user domain disk. I want
> to move it to another slot in ba350 cabinet so I can put a 9gb
> in the internal bay (wide scsi vs narrow...)
>
> Can I just rename the file in the
> /etc/fdmns/domain1:
> area with mv rz1c rz8c
> or whatever the number is?
> and will the files still be intact?
>
Basic solutions involved the Nice way using addvol/rmvol to
migrate disks around or the other way is changing the link
in the /etc/fdmns/domain1
Tom Blinn summarized it pretty nicely so I'll include his
step by step guide... (others also had nice summaries for
the addvol way, if interested let me know... I kept
the for future reference.)
I was able to basically umount the disk, move it to a new slot,
change the link and remount it without losing files. Then
after I got stuff setup on the new 9gb, I dumped it back
and changed the /etc/fstab appropriately (I almost forgot that
part... :) )...
Anyways here's some messages relating to what I did:
> From tpb_at_zk3.dec.com Thu Mar 5 14:31:55 1998
> Date: Wed, 04 Mar 1998 08:38:23 -0500
> From: "Dr. Tom Blinn, 603-884-0646" <tpb_at_zk3.dec.com>
> To: "Pam Woods, Systems Manager" <axsymgr_at_UAA.ALASKA.EDU>
> Subject: Re: ADVfs disk swap/change
>
> If you are going to move a disk that contains an AdvFS domain to a new SCSI
> address, and the entire domain is contained on that one disk, and it's the
> only thing on the disk, then the steps are these:
>
> 1) Unmount all the filesets in the domain.
>
> 2) Move the disk to its new address; in your case, you should be able to
> just physically unplug the cartridge from the current slot and plug it in
> in the new location without even shutting down the system. You might do a
> "scu -f /dev/rrz1a stop" first to spin it down, if you want to be gentle.
>
> 3) Verify that the disk is being seen in the new location; first, make the
> device special names if you don't already have them (cd /dev ; ./MAKEDEV
> rz9), then use "scu" and/or "file" to examine the raw device at the "a" or
> "c" partition name (e.g., file /dev/rrz9a or scu -f /dev/rrz9a show dev or
> show capacity) and make sure the device is visible to system software.
>
>4) cd into the existing /etc/fdmns domain directory (/etc/fdmns/domain1)
> and remove the old symbolic link (rm rz1c), then make a new symbolic link
> to the block name for the domain partition (ln -s /dev/rz9c) and you can
> then mount the filesets (mount domain1#fileset1 /domain1/fileset1 or how
> you do it as the case may be).
>
> If you want to put a new disk where the old one was, plug it in, use the scu
> or file commands to make sure it's being seen. scu is better, because file
> just looks at kernel data structures that don't get updated unless you use
> scu to tell the kernel to re-scan the devices on the bus, and that can be a
> nasty thing to do to a running system.
>
> Once the new disk is visible to system software, you can edit the label (or
> put one on if it doesn't already have one) to set up the partitions the way
> you want them (e.g., split the disk 50-50 between, say, the "d" and "e" or
> the "g" and "h" partitions, but don't change the "a" or "c" partitions if
> you are smart, at least, make sure they both start at the beginning of the
> disk, by convention; you might make "a" and "b" split the disk without any
> problems, just leave "a" at the beginning).
>
> If you want to put AdvFS domains on the disk, then use mkfdmn to do it, as
> you would normally, or do whatever else you need to do with the partitions.
>
> Tom
>
> Dr. Thomas P. Blinn, UNIX Software Group, Digital Equipment Corporation
> 110 Spit Brook Road, MS ZKO3-2/U20 Nashua, New Hampshire 03062-2698
> Technology Partnership Engineering Phone: (603) 884-0646
> Internet: tpb_at_zk3.dec.com Digital's Easynet: alpha::tpb
> ACM Member: tpblinn_at_acm.org PC_at_Home: tom_at_felines.mv.net
>
> From tpb_at_zk3.dec.com Thu Mar 5 14:31:55 1998
> Date: Wed, 04 Mar 1998 16:51:22 -0500
> From: "Dr. Tom Blinn, 603-884-0646" <tpb_at_zk3.dec.com>
> To: "Pam Woods, Systems Manager" <axsymgr_at_UAA.ALASKA.EDU>
> Subject: Re: ADVfs disk swap/change
>
>
> > ok thanks. I used your steps and the deed is done.
> >
> > but when I do scu -f /dev/rrz1a show device
> > now: (yes it sees it's a 9gb with show capacity)
> >
> > I see:
> >
> > Warning: Inquiry rz1db-ca and system (rz1db-ca) device names are
> > different.
> >
> > I have not yet mounted/initialized etc - done anything to the disk
> > but plop it in the slot and power back up (wasn't showing as a device
> > at the >>> show device at first.)
> >
> > do I just need to rebuild the kernel to make it happy?
> >
> > ________
> > Pam Woods - axsymgr_at_uaa.alaska.edu
>
> That warning is something you just have to ignore -- the problem is that the
> device is returning a name string that's longer than what DIGITAL UNIX has
> room for in its device name table internally -- so when scu looks at the
> system's internal name for the device, the name string doesn't match what
> the device itself reports. This would be a real problem if the reason the
> kernel's internal name didn't match the device's real name was because the
> device responding wasn't the one the kernel thought it was talking to (for
> instance, if you happened to have two devices set to the same ID but you'd
> had one powered up when the system booted, then powered it off and powered
> up the other one).
>
> Rebuilding the kernel won't change this, unless you rebuilt with the newer
> X5.0 bits (the next major release), where I *think* this is fixed.
>
> Tom
>
I'm also including Kurt's comments about filesets. I decided
to keep it one big partition for a change but make it look like
2 via filesets where in the past I would have kept partitions: a, b, g, h
vs. one big C ... I wanted the oracle separated from the users_1 ...
> From snkac_at_java.sois.alaska.edu Thu Mar 5 14:31:55 1998
> Date: Wed, 04 Mar 1998 15:44:53 -0800
> From: Kurt Carlson <snkac_at_java.sois.alaska.edu>
> To: "Pam Woods, Systems Manager" <axsymgr_at_UAA.ALASKA.EDU>
> Subject: Re: ADVfs disk swap/change
>
>
>
> It's where a domain has multiple mount points... the filesets
> can be moved with advfs utilities to new domains. Effectively,
> you sort of manage them like a partition but you get the benefit
> of advfs utilities for them and you don't have segregated data
> spaces (probably better IO rates on average) and the can share the
> aggregate free space.
>
> Example follows, usr__2f100g has four filesets (mkfset is the magic command).
> Technically it has five, the the tmp there isn't used.
> usr and var are allowed to utilize the full 14gb, the other filesets
> have quotas defined so they can't consume the whole thing.
>
> snkac_at_glacier: df -k | egrep 'Files|usr__2f100g'
> Filesystem 1024-blocks Used Available Capacity Mounted on
> usr__2f100g#usr 14385096 2149719 7092072 24% /usr
> usr__2f100g#var 14385096 2681829 7092072 28% /var
> usr__2f100g#users 2000000 864762 1135238 44% /users
> usr__2f100g#sx 4000000 1493621 2506379 38% /users/sx
>
> snkac_at_glacier: showfsets usr__2f100g
> usr
> Id : 32224c5e.0005d6ef.1.8001
> Files : 42444, SLim= 0, HLim= 0
> Blocks (512) : 4299438, SLim= 0, HLim= 0
> Quota Status : user=off group=off
>
> tmp
> Id : 32224c5e.0005d6ef.3.8001
> Files : 0, SLim= 0, HLim= 0
> Blocks (512) : 0, SLim= 6000000, HLim= 6000000
> Quota Status : user=off group=off
>
> users
> Id : 32224c5e.0005d6ef.4.8001
> Files : 21547, SLim= 0, HLim= 0
> Blocks (512) : 1729524, SLim= 4000000, HLim= 4000000
> Quota Status : user=off group=off
>
> sx
> Id : 32224c5e.0005d6ef.5.8001
> Files : 17987, SLim= 0, HLim= 0
> Blocks (512) : 2987240, SLim= 8000000, HLim= 8000000
> Quota Status : user=off group=off
>
> var
> Id : 32224c5e.0005d6ef.6.8001
> Files : 4883, SLim= 0, HLim= 0
> Blocks (512) : 5363658, SLim= 0, HLim= 0
> Quota Status : user=off group=off
>
> snkac_at_glacier: showfdmn usr__2f100g
>
> Id Date Created LogPgs Domain Name
> 32224c5e.0005d6ef Mon Aug 26 17:16:14 1996 512 usr__2f100g
>
> Vol 512-Blks Free % Used Cmode Rblks Wblks Vol Name
> 1L 28770200 14184112 51% on 128 128 /dev/rz25g
>
> snkac_at_glacier: df -k /tmp # the real /tmp...
> Filesystem 1024-blocks Used Available Capacity Mounted on
> tmp__2f100b#tmp 6291456 303830 5976616 5% /tmp
> snkac_at_glacier: showfdmn tmp__2f100b
>
> Id Date Created LogPgs Domain Name
> 3409ac4c.000ad780 Sun Aug 31 09:39:24 1997 512 tmp__2f100b
>
> Vol 512-Blks Free % Used Cmode Rblks Wblks Vol Name
> 1L 12582912 11953232 5% on 128 128 /dev/rz25b
>
> rz25 has a, b, g partitions in use... b used to be swap, when swap
> was moved off it was taken over by /tmp. a remains an alternate
> root (/) partition, i don't know if they're cloning it or not. dealing
> with / and swap are two of the remaining really good excuses for physical
> partitioning. k
>
>
And you can use the chfsets -b 4000000 -B 4000000 domain3 users1
to set a logical limit to prevent the users from using the whole
disk.. but I did find out that after a vdump/vrestore I had
to reset the limits ...
but now if the users do need more space, I just increase the limit.
Received on Fri Mar 06 1998 - 01:03:41 NZDT