Why does a Solaris install to a disk with valid, pre-existing fdisk partitions sometimes fail?
There is a well known bug that sometimes prevents Solaris frominstalling into an existing partition. Its cause has never been
identified, or its existence officially acknowledged by filling out a
bug report. It is secretly well known only to Sun's Installation
Support team in Chelmsford, MA., who claim that the workaround is
apparent from the message "slice extends beyond end of disk".
I agree that the workaround is simple, but I think some kind of document
explaining the workaround should be returned by searches of sunsolve and
access1. Better yet, the error message could actually describe the
error! Or, how about identifying and fixing the bug so it never happens
to begin with?
For those of you too "stupid" :-) to read the error message, I'll decode it:
slice = "disk"
extends = "is full of fdisk partitions"
beyond = "before"
end = "installation."
of = "Please"
disk = "delete at least one of 'em, and try again"
For example: If a disk has three partitions with the following:
1) FAT,
2) no filesystem yet,
3) NTFS, the installation might fail
in some poorly understood cases, with the misleading error message.
The workaround is to delete the unused partition, leaving a "hole"
between the flanking partitions. The install fdisk, Partition Magic, or
any other fdisk will now see only two partitions: FAT and NTFS.
There will obviously be lots of cylinders between the end of the first,
and the beginning of the second. The Solaris install will spot the hole,
and create a partition according to its own mysterious specifications.
Somehow, this new partition is acceptable, even though a seemingly
identical one created by a different fdisk isn't. Perhaps the bug is in
*when* it was created: if previous, sometimes balk. Perhaps NORMAL /
LBA is relevant at this point - - it did make a difference in at least
one case I tested. Oddly, I have also had cases where the offending
procedure of creating the partitions before beginning to install Solaris
worked fine.
However, I once had a case where the Solaris install created a partition
that left gaps of a few cylinders before and after. I am wary that it
could err on the other side of the boundary, and damage a flanking
filesystem by encroaching across the pre-existing boundary. The safest
approach when dealing with a squirrelly fdisk is to use the dangerous
one *first*. Install Solaris before the other partitions get used, if
possible. Hopefully the other fdisk-type programs will recognize such
corruption and allow the encroached-upon partitions to be deleted and
re-created, without hurting the Solaris partition.
At any rate, the most-likely-to-succeed procedure is to install into a
hole, or onto an empty disk with no partitions.
Thanks to Super-User (asianinter.net), who pointed out cases involving
modern BIOS' auto-detecting ATAPI disks as NORMAL. Alan Thomas prefers
always to set disks to NORMAL, and once had trouble with a disk that was
set to LBA.
[Thanks to Randy J. Parker]
| Home | FAQ |
XML Feeds