When you mean with “overlapping” that the end address of the first is the same as the start address of the second, I can tell you that I haveseen this lately in several fdisk -l outputs here on the forums. I also was embarrased by it, but those whoo popsted this had no problems with it at all (they osted for different items).
The only thing I can think of that the border is in fact somewhere inside a cylinder. IIRC those cylinders are a fake of the disk firmware anyway.
I like fdisk too, however, I’d like to duplicate the layout and install of this particular disk, but cannot with just fdisk it seems, as it rounds up to the next cylinder. Is there some partitioning (linux based) tool that can take an exact copy of a partition layout and setup another disk (identical) with that identical partition scheme? parted has the tantalizing -m option to make its output more machine readable, but maddeningly doesn’t seem to have a corresponding -r to read a layout that it has generated. I suppose I could script up something to parse the machine output and have it generate the right commands for parted to generate the identical layout…
A quick note: copying the MBR+partition table using dd bs=512 count=1, apparently does not quite work, as parted apparently finds something wrong with it, not sure what parted is looking for (perhaps it is trying to interpret something about the file systems on the partitions?), but it means the GUI partition tool is worthless on a disk with a cloned partition table like this. I’m beginning to suspect there is something going on with the underlying hardware (PERC 5i to SAS disks). I punted by using the GUI partitioning tool with the “Expert” option of Clone this Disk, which apparently just clones the partition information, but no the data on disk…
On 2010-08-05 05:36, jcolbyk wrote:
> After a fresh install of opensuse 11.3 x86_64, using a NET install CD, I
> noticed that the boot disk layout has overlapping partitions:
> atadb-2:~ # fdisk -l /dev/sda
> Disk /dev/sda: 146.2 GB, 146163105792 bytes
> 255 heads, 63 sectors/track, 17769 cylinders
> Units = cylinders of 16065 * 512 = 8225280 bytes
> Sector size (logical/physical): 512 bytes / 512 bytes
> I/O size (minimum/optimal): 512 bytes / 512 bytes
> Disk identifier: 0x000c9ce8
> Device Boot Start End Blocks Id System
> /dev/sda1 1 262 2103296 82 Linux swap / Solaris
> Partition 1 does not end on cylinder boundary.
> /dev/sda2 * 262 2873 20972544 83 Linux
> /dev/sda3 2873 17770 119660544 83 Linux
> I’ve noticed one other post that mentioned this at the very end. Is
> this a known problem already? Or is there something I’m missing that
> makes this okay?
It is OK, AFAIK. Some system insisted on partitions ending on a track end, but this should not be
necessary, we use LBA addressing after all, no?
Change “units” in fdisk and display the table again. You will see then the real limits. Try with
other partitioners and compare their outputs.
I like printing that info. Then the paper gets lost, of course…
Cheers / Saludos,
Carlos E. R.
(from 11.2 x86_64 “Emerald” GM (Elessar))
Regarding the OP, I’m seeing the same thing over and over as well with 11.3 in both 32 and 64 bit fresh installs. Some systems won’t boot after install (error loading operating system") other systems seem fine despite the “overlap”. I doin’t know “for sure” that it’s the “overlap” but on the systems that wouldn’t boot it seems I’m consistently able to fix “the problem” by fdisking the partitions myself from an alt-console as early as possible in the install process and then just picking them during partition setup, using setup to format them and pick mount point but not any of the partitioning. Maybe it’s just dumb luck, but this seems to fix those systems that were getting “error loading operating system” so I’ve come to do all my 11.3 installs this way now as standard practice to consistently ensure it’s not an issue.
I lack the technical knowledge in this particular area to say exactly what’s up, but it seems to me, anecdotally and experientially at least, that something is awry in the 11.3 partition setup tool. Some setups tolerate it and others don’t for whatever reason, but I’d love to see this explained.
Is it really an issue or is there a bug to be squished here?
[compat] partition_alignment added to general section (bnc#597723). On openSUSE 11.3, parted is not creating partitions on cylinder boundaries anymore. That can lead to boot problems on some hardware. See documentation of the general section
(BTW: while following the link to the bugzilla I got an
You are not authorized to access bug #597723.
have we got some kind of “secret” bug here?
Following the last link I find a
<!-- the storage section was invented with openSUSE 11.3 (not SLES11 SP1) -->
`align_optimal - That's the default. Partitions are aligned like the kernel suggests.
This can lead to problem with some machines/bioses that are unable to boot with that
`align_cylinder - that's the alignment like it was in pre-openSUSE 11.3 time for years. Partitions
always start on a cylinder boundary
I will give this a try and if it works I’ll report it here. When this is the solution (at least for autoyast) I kindly request to change the partion_alignment param to align_cylinder by default.
@gogalthorp: I know I should not expect that my request gets an official followup, however, dev’s tend to read the forum too.
anyway, bad news on testing the autoyast.xml!
the allign cylinder change did not solve my problem, however, I did not have enough time to check the partitiontable, just started an automated installation with the autoyast as below and ran into a “could not mount partition” error upon load of grub…
Any comments on my autoyast are much welcome. I will try to contact Uwe as well by the way, keep you posted!