* Internal bitmap always 1/1
@ 2011-01-04 16:51 Andrea Gelmini
2011-01-04 17:20 ` John Robinson
2011-01-04 20:28 ` NeilBrown
0 siblings, 2 replies; 3+ messages in thread
From: Andrea Gelmini @ 2011-01-04 16:51 UTC (permalink / raw)
To: linux-raid; +Cc: Neil Brown
Hi all,
and thanks a lot for your effort on MdAdm.
With latest git version of mdadm tools¹ I have
problems with creation of internal bitmap. Doing something like this:
./mdadm -Ss;./mdadm --create /dev/md0 --level=1 --raid-devices=2
--assume-clean --bitmap=internal /dev/sdb5 missing --auto=md -e
1.2;cat /proc/mdstat
mdadm: stopped /dev/md0
mdadm: /dev/sdb5 appears to be part of a raid array:
level=raid1 devices=2 ctime=Tue Jan 4 17:29:06 2011
Continue creating array? y
mdadm: array /dev/md0 started.
Personalities : [linear] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid1 sdb5[0]
55777208 blocks super 1.2 [2/1] [U_]
bitmap: 1/1 pages [4KB], 65536KB chunk
I always have 1/1 pages (also with different partition sizes: 1G
as 80G, no differences).
I bisected (because official Ubuntu 10.10 MdAdm work well), and
I've got this:
b8ab2a50aba72c7316d92d0fb414167844673136 is the first bad commit
commit b8ab2a50aba72c7316d92d0fb414167844673136
Author: NeilBrown <neilb@suse.de>
Date: Tue Nov 17 13:15:34 2009 +1100
Set default bitmap-chunksize for internal bitmaps to at least 64Meg
A small bitmap-chunksize hurts performance without helping
resync speed much - particularly on internal bitmaps.
So set the default to at least 64Meg.
Signed-off-by: NeilBrown <neilb@suse.de>
:100644 100644 fb20383a2af65b778da3653ff1b14cbd32cbe320
3b6ffa041b0f83f0dcd95e455194d0bc7c249602 M mdadm.8
:100644 100644 f9e6a6db262135476cfed766f15bd0f4ec360847
955e832dbb977d432f4fbf90c619a2dbb171707d M super0.c
:100644 100644 a21e7e3baf8eeeeaf8b9f6a71ed6a833937e846e
8678e9f421b9f45eebaf60849729acf471ad7039 M super1.c
:040000 040000 c38915f28a038db7bcf2a207f3b60731d5655bea
fb66f4f76ef1ff31a00aecd69c1232beee15d82e M tests
Here the bisect log:
gelma@移動:/tmp/mdadm$ git bisect log
git bisect start
# bad: [a2973b6af235f993b3085cfb51bc50cc6ca38d38] segfault in imsm
create with wrong arguments
git bisect bad a2973b6af235f993b3085cfb51bc50cc6ca38d38
# good: [c04bf2398e49763460c0e04df3130566ced16555] Release 2.6.7.1
git bisect good c04bf2398e49763460c0e04df3130566ced16555
# good: [866d136a2f75b751188dfdd490be4d64e6a14d09] Release 2.6.7
git bisect good 866d136a2f75b751188dfdd490be4d64e6a14d09
# good: [6c40598f598874d1d4c2c4d0da0c2a9b873d768d] Merge branch
'master' into devel-3.0
git bisect good 6c40598f598874d1d4c2c4d0da0c2a9b873d768d
# good: [4a997737a14f580e3addfa1db6be005c37e7c0f4] Merge branch
'master' into devel-3.1
git bisect good 4a997737a14f580e3addfa1db6be005c37e7c0f4
# bad: [ed8fa52f31521b8314859f290c464f77359933a0] mdmon: check select
a writable pid_dir
git bisect bad ed8fa52f31521b8314859f290c464f77359933a0
# skip: [dd9bb2fbed335f69225ea1a3514bd27422288a18] imsm: prune dead
code in validate_geometry_imsm
git bisect skip dd9bb2fbed335f69225ea1a3514bd27422288a18
# bad: [c1e3ab8c1e76f4ae71ab23bcf5e8c2bf8bd3774e] Merge branch
'master' of git://github.com/djbw/mdadm
git bisect bad c1e3ab8c1e76f4ae71ab23bcf5e8c2bf8bd3774e
# bad: [cd77ac4eafb0460ff6229e9a8dd732f2c4dd42c5] Assemble: fix
testing of 'verbose' flag.
git bisect bad cd77ac4eafb0460ff6229e9a8dd732f2c4dd42c5
# good: [14e5b4d72b4db55e688581d98ec47131554f747c] Grow: data_offset
is in sectors, offsets[] is in bytes - convert
git bisect good 14e5b4d72b4db55e688581d98ec47131554f747c
# good: [7d5c3964ccfaace123f7b75e15d38c2650e013d8] Change default
metadata from 0.90 to 1.1
git bisect good 7d5c3964ccfaace123f7b75e15d38c2650e013d8
# bad: [b42f577a0dbe0b232600478f4a5068b2f63a2684] Improve error
messages when metadata handler does not support request.
git bisect bad b42f577a0dbe0b232600478f4a5068b2f63a2684
# bad: [b8ab2a50aba72c7316d92d0fb414167844673136] Set default
bitmap-chunksize for internal bitmaps to at least 64Meg
git bisect bad b8ab2a50aba72c7316d92d0fb414167844673136
# good: [ff94fb86fd22842c06e8fad08c27a4e365d7834a] Grow: various fixes
to recent breakages.
git bisect good ff94fb86fd22842c06e8fad08c27a4e365d7834a
Thanks a lot for your work,
Andrea
-----------------------------
¹ git://neil.brown.name/mdadm a2973b6af235f993b3085cfb51bc50cc6ca38d38
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: Internal bitmap always 1/1
2011-01-04 16:51 Internal bitmap always 1/1 Andrea Gelmini
@ 2011-01-04 17:20 ` John Robinson
2011-01-04 20:28 ` NeilBrown
1 sibling, 0 replies; 3+ messages in thread
From: John Robinson @ 2011-01-04 17:20 UTC (permalink / raw)
To: Andrea Gelmini; +Cc: linux-raid
On 04/01/2011 16:51, Andrea Gelmini wrote:
> Hi all,
> and thanks a lot for your effort on MdAdm.
> With latest git version of mdadm tools¹ I have
> problems with creation of internal bitmap. Doing something like this:
> ./mdadm -Ss;./mdadm --create /dev/md0 --level=1 --raid-devices=2
> --assume-clean --bitmap=internal /dev/sdb5 missing --auto=md -e
> 1.2;cat /proc/mdstat
> mdadm: stopped /dev/md0
> mdadm: /dev/sdb5 appears to be part of a raid array:
> level=raid1 devices=2 ctime=Tue Jan 4 17:29:06 2011
> Continue creating array? y
> mdadm: array /dev/md0 started.
> Personalities : [linear] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
> md0 : active raid1 sdb5[0]
> 55777208 blocks super 1.2 [2/1] [U_]
> bitmap: 1/1 pages [4KB], 65536KB chunk
>
> I always have 1/1 pages (also with different partition sizes: 1G
> as 80G, no differences).
Why is this a problem? Your bitmap is simply too small to need more than
one page, whether it's a 1G or 80G. With a 16MB bitmap chunk size on a
2TB array I have 2 pages.
If you want to see how busy the bitmap is, try `mdadm --examine-bitmap
/dev/component`
Cheers,
John.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Internal bitmap always 1/1
2011-01-04 16:51 Internal bitmap always 1/1 Andrea Gelmini
2011-01-04 17:20 ` John Robinson
@ 2011-01-04 20:28 ` NeilBrown
1 sibling, 0 replies; 3+ messages in thread
From: NeilBrown @ 2011-01-04 20:28 UTC (permalink / raw)
To: Andrea Gelmini; +Cc: linux-raid
On Tue, 4 Jan 2011 17:51:17 +0100 Andrea Gelmini <andrea.gelmini@gmail.com>
wrote:
> Hi all,
> and thanks a lot for your effort on MdAdm.
> With latest git version of mdadm tools¹ I have
> problems with creation of internal bitmap. Doing something like this:
> ./mdadm -Ss;./mdadm --create /dev/md0 --level=1 --raid-devices=2
> --assume-clean --bitmap=internal /dev/sdb5 missing --auto=md -e
> 1.2;cat /proc/mdstat
> mdadm: stopped /dev/md0
> mdadm: /dev/sdb5 appears to be part of a raid array:
> level=raid1 devices=2 ctime=Tue Jan 4 17:29:06 2011
> Continue creating array? y
> mdadm: array /dev/md0 started.
> Personalities : [linear] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
> md0 : active raid1 sdb5[0]
> 55777208 blocks super 1.2 [2/1] [U_]
> bitmap: 1/1 pages [4KB], 65536KB chunk
>
> I always have 1/1 pages (also with different partition sizes: 1G
> as 80G, no differences).
Why do you think that this is a problem???
NeilBrown
>
> I bisected (because official Ubuntu 10.10 MdAdm work well), and
> I've got this:
>
> b8ab2a50aba72c7316d92d0fb414167844673136 is the first bad commit
> commit b8ab2a50aba72c7316d92d0fb414167844673136
> Author: NeilBrown <neilb@suse.de>
> Date: Tue Nov 17 13:15:34 2009 +1100
>
> Set default bitmap-chunksize for internal bitmaps to at least 64Meg
>
> A small bitmap-chunksize hurts performance without helping
> resync speed much - particularly on internal bitmaps.
>
> So set the default to at least 64Meg.
>
> Signed-off-by: NeilBrown <neilb@suse.de>
>
> :100644 100644 fb20383a2af65b778da3653ff1b14cbd32cbe320
> 3b6ffa041b0f83f0dcd95e455194d0bc7c249602 M mdadm.8
> :100644 100644 f9e6a6db262135476cfed766f15bd0f4ec360847
> 955e832dbb977d432f4fbf90c619a2dbb171707d M super0.c
> :100644 100644 a21e7e3baf8eeeeaf8b9f6a71ed6a833937e846e
> 8678e9f421b9f45eebaf60849729acf471ad7039 M super1.c
> :040000 040000 c38915f28a038db7bcf2a207f3b60731d5655bea
> fb66f4f76ef1ff31a00aecd69c1232beee15d82e M tests
>
> Here the bisect log:
> gelma@移動:/tmp/mdadm$ git bisect log
> git bisect start
> # bad: [a2973b6af235f993b3085cfb51bc50cc6ca38d38] segfault in imsm
> create with wrong arguments
> git bisect bad a2973b6af235f993b3085cfb51bc50cc6ca38d38
> # good: [c04bf2398e49763460c0e04df3130566ced16555] Release 2.6.7.1
> git bisect good c04bf2398e49763460c0e04df3130566ced16555
> # good: [866d136a2f75b751188dfdd490be4d64e6a14d09] Release 2.6.7
> git bisect good 866d136a2f75b751188dfdd490be4d64e6a14d09
> # good: [6c40598f598874d1d4c2c4d0da0c2a9b873d768d] Merge branch
> 'master' into devel-3.0
> git bisect good 6c40598f598874d1d4c2c4d0da0c2a9b873d768d
> # good: [4a997737a14f580e3addfa1db6be005c37e7c0f4] Merge branch
> 'master' into devel-3.1
> git bisect good 4a997737a14f580e3addfa1db6be005c37e7c0f4
> # bad: [ed8fa52f31521b8314859f290c464f77359933a0] mdmon: check select
> a writable pid_dir
> git bisect bad ed8fa52f31521b8314859f290c464f77359933a0
> # skip: [dd9bb2fbed335f69225ea1a3514bd27422288a18] imsm: prune dead
> code in validate_geometry_imsm
> git bisect skip dd9bb2fbed335f69225ea1a3514bd27422288a18
> # bad: [c1e3ab8c1e76f4ae71ab23bcf5e8c2bf8bd3774e] Merge branch
> 'master' of git://github.com/djbw/mdadm
> git bisect bad c1e3ab8c1e76f4ae71ab23bcf5e8c2bf8bd3774e
> # bad: [cd77ac4eafb0460ff6229e9a8dd732f2c4dd42c5] Assemble: fix
> testing of 'verbose' flag.
> git bisect bad cd77ac4eafb0460ff6229e9a8dd732f2c4dd42c5
> # good: [14e5b4d72b4db55e688581d98ec47131554f747c] Grow: data_offset
> is in sectors, offsets[] is in bytes - convert
> git bisect good 14e5b4d72b4db55e688581d98ec47131554f747c
> # good: [7d5c3964ccfaace123f7b75e15d38c2650e013d8] Change default
> metadata from 0.90 to 1.1
> git bisect good 7d5c3964ccfaace123f7b75e15d38c2650e013d8
> # bad: [b42f577a0dbe0b232600478f4a5068b2f63a2684] Improve error
> messages when metadata handler does not support request.
> git bisect bad b42f577a0dbe0b232600478f4a5068b2f63a2684
> # bad: [b8ab2a50aba72c7316d92d0fb414167844673136] Set default
> bitmap-chunksize for internal bitmaps to at least 64Meg
> git bisect bad b8ab2a50aba72c7316d92d0fb414167844673136
> # good: [ff94fb86fd22842c06e8fad08c27a4e365d7834a] Grow: various fixes
> to recent breakages.
> git bisect good ff94fb86fd22842c06e8fad08c27a4e365d7834a
>
> Thanks a lot for your work,
> Andrea
>
> -----------------------------
> ¹ git://neil.brown.name/mdadm a2973b6af235f993b3085cfb51bc50cc6ca38d38
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2011-01-04 20:28 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-01-04 16:51 Internal bitmap always 1/1 Andrea Gelmini
2011-01-04 17:20 ` John Robinson
2011-01-04 20:28 ` NeilBrown
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).