CN.RNA@forsythe.stanford.edu (Richard Anderson) (03/22/90)
I have an RT 6150 with two 70E hard disks. I am running AOS 4.3.
Since it is for me a new computer, I initially was only using the
first hard disk (hd0). Once I had mastered essential stuff (like
making backups), I moved on to the creation of a new file systems on
the second hard disk (hd1).
Since I am mostly teaching myself Unix system administration with
the help of a few good books, I have been prone to trial and error
experimentation. During this experimentation, the contents of the
filesystem stored on hd1g mysteriously dissapeared. Also, I kept
getting an fsck error message I had trouble deciphering that read:
SUMMARY INFORMATION BAD (SALVAGED)
I suspect the cause of my trouble was due to the fact that I
partioned hd1 into sizes that were both non-standard and
different from the partions sizes of hd0.
I used the minidisk command to set up the first disk as a standard
root drive. hd0a is root, hd0b is swap space, hd0g holds /usr.
I set up the second disk with standard root partitions also, but
then deleted and re-created hd1b and hd1g, so I could create a
larger hd1g. I have used hd1a for a dd backup of hd0a. I also
found out by trial and error that hd1b had to be a swap partition
if I ever wanted to boot from hd1 (if I ever need to repair hd0a).
standard root
delete hd1b
delete hd1g
create hd1b 32743 11025 swap
create hd1g 32744 97735 00
512 Byte Blocks
---------------
hd0 standard root name iodn size start type
------ ------ ------ ------- ------
boot 32736 105 140 ipl
hd0a 32737 32585 245 00
hd0b 32738 33565 32830 swap
hd0g 32739 75195 66395 00
hd1 customized name iodn size start type
------ ------ ------ ------- ------
boot 32741 105 140 ipl
hd1a 32742 32585 245 00
hd1b 32743 11025 32830 swap
hd1g 32744 97735 43855 00
The above partition structure for hd1 did not work properly.
It appears that some parts of the Unix software recognize the sizes
of the partitions, whereas other parts (perhaps the swapper daemon)
thought that both disks were partitioned the same. Whatever the
cause, however, what happened is that all the subdirectories and
files on hd1g were mysteriously deleted at some point from one day
to the next. I cannot rule out human error, but it seems most
likely that the swapper daemon was at fault. Hardware diagnostics
were run by an IBM service rep, but no defects were found.
hd1 was therefore re-partitioned to the standard root configuration.
Both hard drives now have exactly the same partition sizes.
So far, so good. No data has been lost from hd1g since the
re-partitioning.
So, does anybody have any suggestion/advice to offer? Should I have
trusted the author of one of my books who warned me to set up both
my disks the same? If so, why is there a "standard source" option
under minidisk? Does the very existence of hd1b cause the system
to want to use it for swap space, and if so, does the swapper assume
that hd0b and hd1b are the same size? Did my use of dd cause any
problems?
Sorry for the length of my query. Hope the feedback I get will be
educational for the newsgroup. Thanks.
-- Richard Anderson cn.rna@forsythe.stanford.edu
Systems Department, Green Library
Stanford, CA 94305-6004 (415) 725-7932edler@jan.ultra.nyu.edu (Jan Edler) (03/22/90)
In article <8603@lindy.Stanford.EDU> CN.RNA@forsythe.stanford.edu
(Richard Anderson) writes of problems with a 2-drive system in which he
has a spare copy of the root partition on the second drive, and has
trouble because his second drive doesn't also have a swap partition the
same size as that on the first drive.
Here at NYU we've never felt the need to keep a spare root partition on
the second (or third) drive of a machine. There are several reasons
for this:
1) A machine with only 70MB drives doesn't really have a lot of space
to waste,
2) We have a bunch of RTs, so our resources available to help a suffering
machine are generally adequate. E.g. we don't even backup our
root and /usr partitions, since all the machines have essentially
the same stuff on them; the rest can be re-created if necessary.
3) We use the miniroot diskette when we can't use the regular root
partition. Often, we modify a copy of the ibm-supplied miniroot,
to get it to do what we want. E.g. it is easy to make a miniroot
that can be used for rrestore.
4) Our experience with disk reliability has been very good.
In particular, the only times we've had to rely on 2) or 3)
are when doing things like repartitioning a drive, setting
up a new machine, or upgrading to a new release of the system.
You could probably figure out why your system had trouble with the
configuration you were attempting, but I'd recommend you just take
dumps of your root partition every so often, and restore from it if
you ever need to. It really is much simpler, and you'll hardly
ever have to do it. You can certainly partition the second drive
any way you want to.
Jan Edler
NYU Ultracomputer Research Laboratory
edler@nyu.eduwebb@bass.tcspa.ibm.com (Bill Webb) (03/28/90)
>... > hd1 was therefore re-partitioned to the standard root configuration. > Both hard drives now have exactly the same partition sizes. > So far, so good. No data has been lost from hd1g since the > re-partitioning. > > So, does anybody have any suggestion/advice to offer? Should I have > trusted the author of one of my books who warned me to set up both > my disks the same? If so, why is there a "standard source" option > under minidisk? Does the very existence of hd1b cause the system > to want to use it for swap space, and if so, does the swapper assume > that hd0b and hd1b are the same size? Did my use of dd cause any > problems? > I don't have any constructive suggestions but I hope I can at least answer some of your questions. There should be no need to have all swap partitions the same size, though the original BSD 4.2 and 4.3 documentation assumed that they were (since the partition sizes were hard coded into the driver). We (IBM) added the minidisk stuff so that partition sizes could be changed and for compatibility and coexistance with AIX. If the sizes are wildly different then you tend to waste some kernel resources in the maps that keep track of swap space utilization and don't spread the paging evenly over several drives but these are secondary effects. In order to use more than one swap partition it must both be configured into your kernel and you must start swapping on it with a swapon (there is a swapon -a in /etc/rc so all that is really required in addition to the kernel /dev/hd1b /swap swap sw 0 0 in /etc/fstab. Until a swapon is done you only swap on the first swap device configured into your kernel. One machine I looked at here used hd0b, hd1b, hd2b, and sc0b as swap devices, they had 19355, 45080, 152615, and 66896 blocks of swap space respectively (that was the machine we used to test to see if we could malloc 128MB if I remember correctly). This system does not have problems with trashing filesystems. Your use of dd should not have cause problems as the kernel will check to make sure that you don't go outside of a partition. I cannot explain your problems unless it is possible that you failed to reboot the machine after changing the partitions with minidisk. If you didn't reboot then you could definitely could have had problems as the new partitions don't take effect until after a reboot and mkfs and newfs would have affected the old partitions only. Even then though I don't think you would have gotten the symptoms that you did, so I'm at a loss to explain what happened. Since you've since gotten it to work by repartitioning we may never know what the actual problem was. > Sorry for the length of my query. Hope the feedback I get will be > educational for the newsgroup. Thanks. > > -- Richard Anderson cn.rna@forsythe.stanford.edu > Systems Department, Green Library > Stanford, CA 94305-6004 (415) 725-7932 I hope I've been able to answer at least some of the queries. ---------------------------------------------------------------- The above views are my own, not necessarily those of my employer. Bill Webb (IBM AWD Palo Alto), (415) 855-4457. UUCP: ...!uunet!ibmsupt!webb