Partition geometry anomaly?

[Snowhog]

Used Gparted 7.1-5 to create from the extended partition, the two partitons for Natty. 10.0GB for root and 25.0 GB for /home. These are sda9 and sda10 respectively, and are the last two partitions.

When I created the 10.0GB ext4 partition, Gparted kept a 2Mb unallocated gap between sda8 and sda9. I was allowed to change it to 1Mb, but not to zero. If I changed it to 1MB, it established a 1MB unallocated space at the end of sda9. I left it at 2Mb between sda8 and sda9. This same anomoly was not presented when I created the 25GB ext4 partition. I was allowed to have 0 preceding the start of sda10.

Gparted shows this unallocated space between sda8 and sda9. Within Natty, Partition Manager does not show this 'gap'.

From within Natty, sudo fdisk -l shows:

Disk /dev/sda: 120.0 GB, 120034123776 bytes
255 heads, 63 sectors/track, 14593 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: 0xc528b3c5

  Device Boot   Start     End   Blocks  Id System
/dev/sda1  *      1    1305  10482381  83 Linux
/dev/sda2      1306    14594  106737877+  f W95 Ext'd (LBA)
/dev/sda5      1306    4568  26209984+ 83 Linux
/dev/sda6      4569    4829   2096451  82 Linux swap / Solaris
/dev/sda7      4830    6134  10482381  83 Linux
/dev/sda8      6135    9397  26210016  83 Linux
/dev/sda9      9398    [color=red]10703[/color]  10482688  83 Linux
/dev/sda10     [color=red]10703[/color]    13966  26210304  83 Linux

and sudo fdisk -lu shows:

Disk /dev/sda: 120.0 GB, 120034123776 bytes
255 heads, 63 sectors/track, 14593 cylinders, total 234441648 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0xc528b3c5

  Device Boot   Start     End   Blocks  Id System
/dev/sda1  *     63  20964824  10482381  83 Linux
/dev/sda2    20964949  234440703  106737877+  f W95 Ext'd (LBA)
/dev/sda5    20964951  73384919  26209984+ 83 Linux
/dev/sda6    73384983  77577884   2096451  82 Linux swap / Solaris
/dev/sda7    77577948  98542709  10482381  83 Linux
/dev/sda8    98542773  150962804  26210016  83 Linux
/dev/sda9    150966272  171931647  10482688  83 Linux
/dev/sda10   171933696  224354303  26210304  83 Linux

fdisk -l shows that the End of sda9 and the Start of sda10 share the same cylinder - 10703.
However, fdisk -lu shows that the sector sda9 ends on is not the same sector that sda10 starts on.

Gparted didn't complain, and neither does fdsik or Natty when booted. Should I be worried? Is there anything to be worried about?

[verndog]

Re: Partition geometry anomaly?

What do you use for Align to:

[dibl]

Re: Partition geometry anomaly?

The start and end cylinder should never be the same number. The start cylinder of the following partition should be one cylinder more than the ending cylinder of the preceding partition. For example:

Disk /dev/sda: 16.0 GB, 16013942784 bytes
32 heads, 32 sectors/track, 30544 cylinders, total 31277232 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00042f68

  Device Boot   Start     End   Blocks  Id System
/dev/sda1  *    1024   1047551   523264  83 Linux
/dev/sda2     1047552  31277055  15114752  83 Linux

[Snowhog]

Re: Partition geometry anomaly?

dibl@

I'm aware of that, and so my concern.

[Snowhog]

Re: Partition geometry anomaly?

verndog@

Aligned to MiB (default), as I've always used.

It's interesting that GParted doesn't detect a problem, and, as all my three systems work without any issues, and I never let any of them grow to more than 75% of the partitions size, I don't believe any problems will result.

[oshunluvr]

Re: Partition geometry anomaly?

IMO, partitions may share a cylinder, although it's not desirable. As long as blocks aren't shared, you shouldn't encounter any real problems. I had a similar situation on my server when it was running PCLOS. All was fine until I wanted to install Ubuntu server and Ubiquity didn't like the partitions sharing a cylinder.

Eventually, I messed around enough to inadvertently wipe my partition table and had to spend the better part of a weekend re-creating the table. :P

I suspect somewhere along the partitioning path you followed, you slipped into ext4 "heck" and partition based on size rather than cylinder boundaries and as we all know - size really doesn't matter (sorry, couldn't resist!)

If it were my hard drive, I'd realign the offending partition(s) to exist within the cylinder boundaries and give up the few bytes especially if you're using ext4, which has known problems with totally used hard drives and being mis-aligned with hard drive geometry.

[Snowhog]

Re: Partition geometry anomaly?

Well, all the partitions are ext4. The only way I can eliminate the overlap on the two affected partitions is to resize sda8 and sda9 so that there is a small unallocated gap between sda8/9 and sda9/10.

  Device Boot   Start     End   Blocks  Id System
/dev/sda8      6135    939[color=red]8[/color]  26210725+ 83 Linux
/dev/sda9      939[color=red]8[/color]    1070[color=red]3[/color]  10482688  83 Linux
/dev/sda10     1070[color=red]3[/color]    13966  26210304  83 Linux

Guess that would be the prudent thing to do - eliminates any possibility of trouble later.

[Snowhog]

Re: Partition geometry anomaly?

Fixed.

/dev/sda7      4830    6134  10482381  83 Linux
/dev/sda8      6135    9397  26202533+ 83 Linux
/dev/sda9      9398    10702  10474496  83 Linux
/dev/sda10     10703    13966  26210304  83 Linux

I believe I know how this happened. My HDD is setup for three OS's, with 10GB partitions for root and 25GB partitions for home. There is a single 2GB swap used by each. There's 4.81GiB of remaining unallocated space at the end of the drive. The problem may have resulted because I didn't establish the partitions exactly to size.

The 10GB partitions were sized to 10236MiB, which GParted reports as 10GB, but the actual size should have been 10240MiB.
The 25GB partitions were sized to 25595MiB, which GParted reports as 25GB, but the actual size should have been 25600MiB.

Unless I want to backup the entire HDD - I can - and redo all the partitions correctly then restore from the backup, I can't fix the problem cleanly - without creating a small unallocated gap between sda8/9 and sda9/10. I did however, get these gaps down to the minimum size - 8.94MiB each. I'll live with that for now.

[oshunluvr]

Re: [RESOLVED] Partition geometry anomaly?

I believe you made the prudent choice.

Besides, whats a few MB? One or two photos?

[Snowhog]

Re: [RESOLVED] Partition geometry anomaly?

Likely, this weekend, I'll make a full HDD backup (to my 500GB USB HDD), and redo my HDD and set up the partitions 'correctly' and restore from the backup. Yes, the small 8.94MiB 'gaps' between the affected partitions aren't much, but they do does vex me. So I'll have to remedy it to make me happy.

[Snowhog]

Re: [RESOLVED] Partition geometry anomaly?

Erased my external USB 500GB HDD. Used Parted Magic 5.10 to create the six partitions (ext4) needed on it, then rebooted into Clonezilla 1.2.6-59 and backed up the six partitions on my laptop. That done, I booted back into Parted Magic and deleted all the partitions on my laptop HDD, then reestablished the seven partitions (six plus a swap).

The layout is:

root
home
swap
root
home
root
home

The sizing is:

10GB (10240MiB)
25GB (25600MiB
2GB ( 2048MiB)
10GB (10240MiB)
25GB (25600MiB
10GB (10240MiB)
25GB (25600MiB

Then, while still in Parted Magic, I 'copied' and 'pasted' the backed up partitions on to my laptop HDD. All went without a hitch, and rebooting the PC aftwards I was presented with my Grub2 menu, and all three of my OS's are accessible and all boot and run flawlessly.

But now sudo fdisk -l shows five partitions are sharing cylinders:

Disk /dev/sda: 120.0 GB, 120034123776 bytes
255 heads, 63 sectors/track, 14593 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: 0xc528b3c5

  Device Boot   Start     End   Blocks  Id System
/dev/sda1        1    [b][color=red]1306[/color][/b]  10485760  83 Linux
/dev/sda2      1306    14594  106733568  5 Extended
/dev/sda5      [b][color=red]1306[/color][/b]    [b][color=red]4570[/color][/b]  26214400  83 Linux
/dev/sda6      [b][color=red]4570[/color][/b]    [b][color=red]4831[/color][/b]   2097152  82 Linux swap / Solaris
/dev/sda7      [b][color=red]4831[/color][/b]    6136  10485760  83 Linux
/dev/sda8      6137    [b][color=red]9400[/color][/b]  26214400  83 Linux
/dev/sda9      [b][color=red]9400[/color][/b]    [b][color=red]10706[/color][/b]  10485760  83 Linux
/dev/sda10     [b][color=red]10706[/color][/b]    13969  26214400  83 Linux

But are not sharing sectors:

Disk /dev/sda: 120.0 GB, 120034123776 bytes
255 heads, 63 sectors/track, 14593 cylinders, total 234441648 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0xc528b3c5

  Device Boot   Start     End   Blocks  Id System
/dev/sda1      2048  20973567  10485760  83 Linux
/dev/sda2    20973568  234440703  106733568  5 Extended
/dev/sda5    20975616  73404415  26214400  83 Linux
/dev/sda6    73406464  77600767   2097152  82 Linux swap / Solaris
/dev/sda7    77602816  98574335  10485760  83 Linux
/dev/sda8    98576384  151005183  26214400  83 Linux
/dev/sda9    151007232  171978751  10485760  83 Linux
/dev/sda10   171980800  224409599  26214400  83 Linux

As far as I know, I've never seen this before. But as everything is working properly, and nothing in the boot process complains, I have to chalk up what is being reported by fdisk as an anomoly.

[oshunluvr]

Re: [RESOLVED] Partition geometry anomaly?

(I'm surprised Don hasn't jumped in here yet...)

Well, IMO you're likely gonna be OK unless you run MSDOS 6.22 or a semi-broken disk manager that doesn't like the non-alignment. Windows Vista and 7 formatting tools begin with sector 2048 to force the alignment and fdisk will also if you switch off dos-compatibility mode when you start fdisk. If you do this and then specify sizes in whole MB or larger, you should find cylinder alignment. The way to guarantee cylinder alignment using fdisk is to begin each partition with a sector divisible by 8.

I will say none of my 8 hard drives ended up this way but I also had dos-compat mode turned off.

Device Boot     Start     End   Blocks  Id System
/dev/sda1        1     131   1052226  82 Linux swap / Solaris
/dev/sda2  *     132     156   200812+ 83 Linux
/dev/sda4       157    60801  487130932  5 Extended
/dev/sda5  *     157    1200   8385898+ 83 Linux
/dev/sda6      1201    1593   3156741  fd Linux raid autodetect
/dev/sda7      1594    2115   4192933+ fd Linux raid autodetect
/dev/sda8      2116    7337  41945683+ fd Linux raid autodetect
/dev/sda9      7338    7859   4192933+ fd Linux raid autodetect
/dev/sda10      7860    8643   6297448+ 83 Linux
/dev/sda11      8644    9427   6297448+ 83 Linux
/dev/sda12      9428    10211   6297448+ 83 Linux

[dibl]

Re: [RESOLVED] Partition geometry anomaly?

It's prolly OK, Snowhog -- I think oshunluvr is correct about the concern with MS-DOS only. Hard drive partitions can end in mid-cylinder.

My most recent partitioning adventures have been with SSDs, so I had to learn to use fdisk to align the first "cylinder" of each partition (there are no real cylinders on SSDs). But I don't think Parted will lead you wrong.

[oshunluvr]

Re: [RESOLVED] Partition geometry anomaly?

See, I know I'd get him to speak up

I think this is mostly a product of new drives and old tools that aren't all working the same way yet. Really, the drive control firmware is remapping the drive layout anyway, so we're really just playing with the numbers.

If you were using winders, I might recommend you try again. Based on your format table, it doesn't seem that's going to happen....

[Snowhog]

Re: [RESOLVED] Partition geometry anomaly?

As the partitions aren't 'overlapping' when one looks at them as number of sectors, and none of the formatting tools are reporting a problem with the layout, I have to accept that all is okay (and I am not using DOS anything). Gparted, Parted Magic, and KDE's partitionmanager, all show the disk partitions in sizes I established (10GB, 25GB, and my 2GB swap), so I'll continue to just accept that some annomaly between these three apps and fdsik exists.

I'm only perflexed, because I've never seen this previously. fdisk has always show partitions with no overlap.