LVM Performance effects?

(sorry for disrupting the thread - copied this message from the archives)

Me too.

Have you found an answer to that problem? I couldn't find anything helpful in the archives. Is this a general LVM on software RAID issue?

I'm measuring a similar performance degradation on my system:

plain Software RAID-5:

# hdparm -t /dev/md3

/dev/md3:
Timing buffered disk reads: 884 MB in 3.00 seconds = 294.49 MB/sec

LVM2 logical volume:

# hdparm -t /dev/data/temp

/dev/data/temp:
Timing buffered disk reads: 316 MB in 3.01 seconds = 105.15 MB/sec

I tried changing the physical extent size, but this has almost no effect.

Some short system information, I'll gladly provide more if needed:

- kernel 2.6.17
- lvm2

LVM version: 2.02.06 (2006-05-12)
Library version: 1.02.07 (2006-05-11)
Driver version: 4.5.0

- 2x dual core xeon, 3 GHz
- 6x WD360GD 36,7 GB SATA Disk
- Software RAID-5 (chunk size: 256kB)

--- Physical volume ---

  PV Name               /dev/md3
  VG Name               data
  PV Size               167.45 GB / not usable 0
  Allocatable           yes
  PE Size (KByte)       4096
  Total PE              42868
  Free PE               4468
  Allocated PE          38400

--- Volume group ---
VG Name data
System ID

  Format                lvm2
  Metadata Areas        1
  Metadata Sequence No  11
  VG Access             read/write
  VG Status             resizable
  MAX LV                0
  Cur LV                6
  Open LV               0
  Max PV                0
  Cur PV                1
  Act PV                1
  VG Size               167.45 GB
  PE Size               4.00 MB
  Total PE              42868
  Alloc PE / Size       38400 / 150.00 GB
  Free  PE / Size       4468 / 17.45 GB

--- Logical volume ---

  LV Name                /dev/data/temp
  VG Name                data
  LV Write Access        read/write
  LV Status              available
  # open                 0
  LV Size                10.00 GB
  Current LE             2560
  Segments               1
  Allocation             inherit
  Read ahead sectors     0
  Block device           253:5

Any hints are greatly appreciated.

Thanks,
Michael

/ 2006-07-13 17:56:55 +0200
\ Michael Heyse:

maybe lvm (device mapper) does not propagate the read-ahead parameters of the physical device to the upper layers
(we had a similar problem with drbd, which in respect to this problem is "just an other" virtual block device)

once we did (block device kernel code during device creation) +
+ if( q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) {
+ INFO("Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
+ q->backing_dev_info.ra_pages,
+ b->backing_dev_info.ra_pages);
+ q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages;
+ }
+

we suddenly got basically all the bandwidth the box could deliver. without that patch and that particular storage backend we had only about 1/6 of the performance.
other storage backends don't care at all.

you could verify this with "blockdev --getra /dev/whatever", and try to tune it with "blockdev --setra 1024 /dev/whatever".

: Lars Ellenberg                                  Tel +43-1-8178292-0  :
: LINBIT Information Technologies GmbH            Fax +43-1-8178292-82 :
: Schoenbrunner Str. 244, A-1120 Vienna/Europe   http://www.linbit.com :