On Tue, 16 Sep 2003, James Melin wrote:
> Once upon a time, our former storage administrator carved up a new shark
> box and gave me 8 3390-mod 9's and 8 3390-mod 3's, for 2 Linux Lpars.
> Things went merrily along until we attempted to do a disaster recovery test
> of Linux, where I discovered that the 3390-9's were exactly 100 cylinders
> too large.
>
> So I have mod-9 architected dasd carved up in ESS (shark) box, at 10117
> cyl, as opposed to 10017 cyl. Near as we can tell this was a fat finger
> mistake. This HAS resulted in my being unable to use dfdss tools to restore
> the backup to a real mod 9, because the dataset is too big.
>
> I am fortunate that 1/2 of the aforementioned dasd has not been made into a
> system, and I can re-format those partitions under Linux to be the correct
> size, and then use an internal copy process to copy the data.
>
> The questions I have are this:
My question is, "What filesystem?"
If ext{2,3} then you can resize the space in-place using resize2fs and
then dd. Probablty, you can do it with reiser too (I see a program that
looks like it might do it).
I think that for ext{2,3} you must force e2fsck first - read docs for
both commands.
To find what to resize to, create empty filesystem(s) in the destination
location(s) and see what the utils say.
>
> How many 4096 byte blocks make up 10117 cylinders without going over, and
> allowing room for the VTOC?
>
> Secondly, since DD will likely not work to copy the current images to the
> slightly smaller partitions on the other volumes, does anyone have a
> preferred method of doing a file system copy that will copy everything
> verbatim (keep the user info, group info, symlinks etc) but do it at the
> file level?
>
> I have looked at pax and tar, and they have me confused on the exact syntax
> to accomplish what I want. DD is the most straightforward, but I am
> concerned that using a block level disk copy tool I will lose data or
> otherwise munge things moving to a smaller partition.
>
> The file system utilization is low
>
> Filesystem 1k-blocks Used Available Use% Mounted on
> /dev/dasdb1 7169608 255592 6549812 4% /
> /dev/dasdc1 7169608 40264 6765140 1% /var
> /dev/dasdd1 7169608 4039356 2766048 60% /opt
> /dev/dasde1 7169608 838428 5966976 13% /usr
> /dev/dasdf1 858100 9280 805232 2% /tmp
>
> So, can I get away with a dd copy, then mount everything under /mnt and
> zipl, or does the collected wisdom of the list suggest another method? If
> so what is the syntax that will succeed at doing what DD does, but at the
> file level instead of block level?
>
> -James
>
--
Cheers
John.
Join the "Linux Support by Small Businesses" list at
http://mail.computerdatasafe.com.au/mailman/listinfo/lssb
Copyright John Summerfield. Reproduction prohibited.
