GNU bug report logs - #25103
partition alignment

Previous Next

Full log

View this message in rfc822 format

From: Volta <xvolta <at> gmail.com>
To: 25103 <at> debbugs.gnu.org
Subject: bug#25103: partition alignment
Date: Sun, 04 Dec 2016 01:54:33 +0000

[Message part 1 (text/plain, inline)]

I'm not sure if this is a bug in parted (parted 3.2 on Ubuntu 16.04.1LTS
x86-64), or a bug in my understanding of the concepts.  Either way I'd like
to know.

I've got a new 8TB drive for which the "optimal" alignment calculated by
parted doesn't satisfy the "minimal" alignment check. My understanding is
the "minimal" check verifies the partition is aligned to physical blocks on
the disk, so it seems impossible for an alignment that doesn't pass that
check to truly be "optimal".  Indeed, the documentation specifies "optimal"
is a aligned to a multiple of the physical block size.

For this drive, "--align opt" creates a partition starting at 65535s:

-- 8< --
$ sudo parted --align opt /dev/sdi mkpart 8tb ext4 0% 100% u s p free a min
1 a opt 1
Model: Seagate Expansion Desk (scsi)
Disk /dev/sdi: 15628053167s
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags:

Number  Start         End           Size          File system  Name  Flags
        34s           65534s        65501s        Free Space
 1      65535s        15628000379s  15627934845s  ext4         8tb
        15628000380s  15628053133s  52754s        Free Space

1 not aligned
1 aligned
-- 8< --

So, I fell back to how we used to do this, but the numbers are... weird:

-- 8< --
$ cat /sys/block/sdi/queue/optimal_io_size
33553920
$ cat /sys/block/sdi/queue/minimum_io_size
4096
$ cat /sys/block/sdi/queue/physical_block_size
4096
$ cat /sys/block/sdi/queue/logical_block_size
512
$ cat /sys/block/sdi/alignment_offset
0
-- 8< --

( 33553920 + 0 ) / 4096 = 8191.875, but that's not going to work. Using the
logical block size instead gives the 65535s that parted suggested, but if
my understanding is correct that's not actually a good alignment.

I can get a partition that satisfies both checks by scaling the 65535s out
to align with a 4k block, and starting the partition at 524280s, at the
cost of "wasting" ~300MiB of space:

-- 8< --
$ sudo parted --align opt /dev/sdi mkpart 8tb ext4 524280s 100% u s p free
a min 1 a opt 1
Model: Seagate Expansion Desk (scsi)
Disk /dev/sdi: 15628053167s
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags:

Number  Start         End           Size          File system  Name  Flags
        34s           524279s       524246s       Free Space
 1      524280s       15628000379s  15627476100s  ext4         8tb
        15628000380s  15628053133s  52754s        Free Space

1 aligned
1 aligned
-- 8< --

~300MiB out of 7.3TiB isn't the end of the world, so that's what I've gone
with for now, but I wonder if I'm not overthinking it.  Was parted's
original optimal alignment actually fine and my understanding incorrect? If
not, is there a better way than what I've done to calculate an alignment
that is?

For the near-term there's no data on this device so I'm perfectly happy to
burn it down for testing.

Thanks,

 -V

-- 8< --
$ sudo parted /dev/sdi print unit s print unit chs print
Model: Seagate Expansion Desk (scsi)
Disk /dev/sdi: 8002GB
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags:

Number  Start  End     Size    File system  Name  Flags
 1      268MB  8002GB  8001GB               8tb

Model: Seagate Expansion Desk (scsi)
Disk /dev/sdi: 15628053167s
Sector size (logical/physical): 512B/4096B
Partition Table: gpt
Disk Flags:

Number  Start    End           Size          File system  Name  Flags
 1      524280s  15628000379s  15627476100s               8tb

Model: Seagate Expansion Desk (scsi)
Disk /dev/sdi: 972801,80,61
Sector size (logical/physical): 512B/4096B
BIOS cylinder,head,sector geometry: 972801,255,63.  Each cylinder is 8225kB.
Partition Table: gpt
Disk Flags:

Number  Start      End         File system  Name  Flags
 1      32,161,57  972798,8,5               8tb
-- 8< --

[Message part 2 (text/html, inline)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.