How to safely merge two slices on harddisk?

[Jerry McAllister]

> 
> Hi,
> 
> I have a hard disk, on which I would like to merge two slices
> into one single slice. The disk slices are as follows:

First, in the df output I see below, there is only one slice
showing.   That would be slice 1 of the /dev/ad1 disk eg. /dev/ad1s1.

Second, it would be more useful in this case to post the output
from disklabel because that would show the sector numbers of
the two partitions (f and g) that you are talking about merging.
  Do:    disklabel -r ad1s1

NOTE:  df shows size estimates after all formatting and system reserves
are taken out of the calculation, so it does not show the actual
number of blocks that were used in creating the label for the disk.
It also does not show unmounted partitions such as a possible swap
partition.  

Here is a sample from one of my disks, 
  (minus stock disk info stuff at the top).

Note, where is says '8 partitions' it means you can defince up 
to 8 partitions, not that 8 partitions have been defined.

Also NOTE:   The size and offset values are given in 512 Byte blocks.

8 partitions:
#        size   offset    fstype   [fsize bsize bps/cpg]
  a:  1048576        0    4.2BSD     1024  8192    22   # (Cyl.    0 - 65*)
  c: 35551782        0    unused        0     0         # (Cyl.    0 - 2212*)
  f:  1048576  1048576    4.2BSD     1024  8192    22   # (Cyl.   65*- 130*)
  g:  6291456  2097152    4.2BSD     2048 16384    89   # (Cyl.  130*- 522*)
  h: 27163174  8388608    4.2BSD     2048 16384    89   # (Cyl.  522*- 2212*)


> /dev/ad1s1a    98M    43M    47M    48%    /home/userB
> /dev/ad1s1d    64G    45G    14G    77%    /home/userA
> /dev/ad1s1e   3.0G   2.5G   282M    90%    /home/userC
> /dev/ad1s1f   3.0G   1.0G   1.7G    37%    /usr/ports
> /dev/ad1s1g   3.0G   268M   2.5G    10%    /mnt
> /dev/ad1s1h   295M   295M -23.5M   109%    /diskless_swap
> 
> I want to merge /ad1s1f and /ad1s1g into one 6Gb slice.

My inclination would be to make a good backup of the partitions
that you want to preserve - regardless of what nice tricks you
want to use and whether they should be safe.

Then, when you have good backups done, the best and easiest thing
to do would be to re-partition the whole /dev/ad1s1 slice and then
reload from backups.

But, once you have the backups, you might be able to just use disklabel
to rewrite the label so the sectors in partitions f and g are combined 
in to a single partition f.    

The key thing to look for in the disklabel -r ad1s1 output is if
the f: and g: partitions are truly adjacent and can be a contiguous
single partition.  This is probably, but it is possible that they
are not.  Nothing requires using partition labels in alphabetical order.

So, add up the size fields of all the partitions before the ones in question.
Presuming the above order and typical usage, that would be a: & d:.
If those sizes equal the offset of the f: partition then that part is OK.
Then add in the size of the f: partition to the previous total and if that 
sum equals the offset value of the g: partition reported by disklabel,
then you can be sure they are adjacent and contiguous.  So, if you are
so inclined, you can take a shot at it.

But, if the size sums and offsets indicate that these two are
not really adjacent, then you are stuck with complately re-labeling
the ad1s1 slice.

One improbable thing could come up.  It is possible that there is some
wasted space that was left out between the f: and g: partitions.  If
this is so, your size sums and offset values should tell you that.
If this occurs, you can just roll that in to the combined f: and g: just by
adjusting the size of the new f: partition.  


To try concatenating the two adjacent partitions do:   
  disklabel -e -r ad1s1
(You could try it once first with -n  (eg disklabel -e -r -n ad1s1) just
 to check and see what it would write)

disklabel -d -r ad1s1 will put the label info in to a file of the
exact format to feed back to disklabel.  Then you can just edit it.
When you get out of the file with a write/save it will write the
label back to the disk.   It will use any editor you have designated,
but the default is vi.

Example:
You should see something like (based on wild guesses from your df above -  I 
did not compensate for how much was taken out by newfs making superblocks, 
system reserves, unmounted partitions, etc, so these are not the real numbers):

8 partitions:
#        size     offset    fstype   [fsize bsize bps/cpg]
  a:  200704         0      4.2BSD    1024  8192    22    # (Cyl.    0 - 65*)
  c: 153896960       0      unused       0     0          # (Cyl.    0 - 2212*)
  d: 134217728  200704      4.2BSD    2048 16384    89    # (Cyl.    0-2212*)
  e: 6291456    134418432   4.2BSD    1024  8192    22    # (Cyl.    0-2212*)
  f: 6291456    140709888   4.2BSD    1024  8192    22    # (Cyl.   65*- 130*)
  g: 6291456    147001344   4.2BSD    1024  8192    22    # (Cyl.  130*- 522*)
  h: 604160     153292800   4.2BSD    1024  8192    22    # (Cyl.  522*- 2212*)

Basically you just delete the line labeled g: and combine the size in
to the line for f:   All the other lines would remain the same.
Given the above numbers, the new f: line would look like:

  f: 12582912   140709888   4.2BSD    2048 16384    22    # (Cyl.   65*- 130*)

NOTE: since I have left out adjustments for formatting and possible waste
at the end, you will see different actual numbers in the disklabel output
and you will have to calculate the actual numbers from what you actually 
see in the "disklabel -r ad1s1" as described above.  

In addition, there may be a b: partition that does not show up in 
the df output.

The h: partition should not be affected either if the offsets and
sizes for all the other partitions are kept the same.    

You should not have to worry about the things in the line past
the fstype (4.2BSD) designation.  You most likely want to take
the defaults - whatever disklabel puts in there for you.   I believe
you can even leage it blank.
Since you want the other partitions to be unaffected, you should not 
change it the settings for them.   

The stuff beyond the # is a comment.

After you finish the disklabel, you need to newfs the new f: partition.
If you used the defaults for fsize, etc then use the newfs devaults.
Just do:
          newfs /dev/ad1s1f

Do not do a newfs for any of the other partitions if you are trying
to get away with just combining the f: and g: partitions without redoing
the whole disk.    But, if you do redo the whole disk and reload from
backup, then all partitions (except those designated unused or swap)
will have to have newfs run on them.

Of course, you will also have to edit your /etc/fstab file
to get rid of the mount for the old /dev/ad1s1g since it will
no longer exist.  You can just comment out the line for now.

> 
> The merging should NOT destroy anything in the slices before
> (ad1s1a, d, and e), but destroying the data in the one afer
> (ad1s1h) is no problem.

I think, but do not know for sure (important caveat), that if you
make the a, d and e partitions (and possibly b: swap partition) have 
the exact same offset and size and that you do NOT run newfs on any 
of a, d or e, then they should remain untouched.   But, make sure you 
have the backups.

Good luck,

////jerry

> 
> Is there a way to do this? What is the safest one?
> (without having to backup the whole disk).
> 
> Thanks,
> Rob.
> 
> 
> _______________________________________________
> freebsd-questions at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
> To unsubscribe, send any mail to "freebsd-questions-unsubscribe at freebsd.org"
>