Discussion:
Block pointer rewrite?
(too old to reply)
Roy Sigurd Karlsbakk
2012-01-06 16:53:46 UTC
Permalink
Hi all

Some time back, perhaps 2-3 years ago, I first heard someone mention Block pointer rewrite (BPR) and what could be done with that. Is it, from the Illumos developers point of view, likely that this will ever see the day? As far as I can understand, S11 doesn't have this, but I thought I heard certain parts of it was written.

PS: The reason I think of this, is I have a pool which is rather badly balanced due to a bad combination of little time and (at that moment) little zfs knowledge, so currently two VDEVs are >95% full, and AFAIK there's no way to fix that except (a) reconfigure everything or (b) by magick or BPR, which may be similar :P
--
Vennlige hilsener / Best regards

roy
--
Roy Sigurd Karlsbakk
(+47) 97542685
***@karlsbakk.net
http://blogg.karlsbakk.net/
--
I all pedagogikk er det essensielt at pensum presenteres intelligibelt. Det er et elementært imperativ for alle pedagoger å unngå eksessiv anvendelse av idiomer med fremmed opprinnelse. I de fleste tilfeller eksisterer adekvate og relevante synonymer på norsk.
Matthew Ahrens
2012-01-09 22:55:47 UTC
Permalink
Post by Roy Sigurd Karlsbakk
Post by Roy Sigurd Karlsbakk
Some time back, perhaps 2-3 years ago, I first heard someone mention
Block pointer rewrite (BPR) and what could be done with that. Is it, from
the Illumos developers point of view, likely that this will ever see the
day? As far as I can understand, S11 doesn't have this, but I thought I
heard certain parts of it was written.
I implemented most of BP rewrite several years back, at Sun/Oracle. I
don't know what plans Oracle has for this work, but given its absence in
S11, I wouldn't bank on it being released. There are several obstacles
that they would have overcome. Performance was a big problem -- like with
dedup, we must store a giant table of translations. Also, the code was
didn't layer well; many other features needed to "know about" bprewrite.
Maintaining it would add significant to cost to future projects.
Post by Roy Sigurd Karlsbakk
PS: The reason I think of this, is I have a pool which is rather badly
balanced due to a bad combination of little time and (at that moment)
little zfs knowledge, so currently two VDEVs are >95% full, and AFAIK
there's no way to fix that except (a) reconfigure everything or (b) by
magick or BPR, which may be similar :P
What problem are you having in particular? My colleague George Wilson has
done some work to improve performance on imbalanced vdevs, which may be
sufficient for you. Be sure you are running a build with the fix for
1051<https://www.illumos.org/issues/1051>,
integrated in May 2011.

--matt



-------------------------------------------
illumos-developer
Archives: https://www.listbox.com/member/archive/182179/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-c42b328b
Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&id_secret=21175072-605db409
Powered by Listbox: http://www.listbox.com
Roy Sigurd Karlsbakk
2012-01-13 08:51:45 UTC
Permalink
Post by Roy Sigurd Karlsbakk
PS: The reason I think of this, is I have a pool which is rather badly balanced
due to a bad combination of little time and (at that moment) little zfs knowledge,
so currently two VDEVs are >95% full, and AFAIK there's no way to fix that except
(a) reconfigure everything or (b) by magick or BPR, which may be similar :P
What problem are you having in particular? My colleague George Wilson has done some
work to improve performance on imbalanced vdevs, which may be sufficient for you.
Be sure you are running a build with the fix for 1051 , integrated in May 2011.
The issue we're seeing is rather terrible performance, especially during scrub, on this system. It has sufficient amounts of L2ARC (10x256GB) and SLOG (a 100GB mirror) (see attached zpool iostat -v - the mirror with c14t21d0 and c14t22d0 are the SLOG devs). It's still at oi_148, so perhaps just an upgrade to 151 might help? Or will I need to patch the system with later code?

Vennlige hilsener / Best regards

roy
--
Roy Sigurd Karlsbakk
(+47) 97542685
***@karlsbakk.net
http://blogg.karlsbakk.net/
--
I all pedagogikk er det essensielt at pensum presenteres intelligibelt. Det er et elementÊrt imperativ for alle pedagoger å unngå eksessiv anvendelse av idiomer med fremmed opprinnelse. I de fleste tilfeller eksisterer adekvate og relevante synonymer på norsk.
Bayard G. Bell
2012-01-16 17:03:24 UTC
Permalink
Post by Roy Sigurd Karlsbakk
Post by Roy Sigurd Karlsbakk
PS: The reason I think of this, is I have a pool which is rather badly balanced
due to a bad combination of little time and (at that moment) little zfs knowledge,
so currently two VDEVs are >95% full, and AFAIK there's no way to fix that except
(a) reconfigure everything or (b) by magick or BPR, which may be similar :P
What problem are you having in particular? My colleague George Wilson has done some
work to improve performance on imbalanced vdevs, which may be sufficient for you.
Be sure you are running a build with the fix for 1051 , integrated in May 2011.
The issue we're seeing is rather terrible performance, especially during scrub, on this system. It has sufficient amounts of L2ARC (10x256GB) and SLOG (a 100GB mirror) (see attached zpool iostat -v - the mirror with c14t21d0 and c14t22d0 are the SLOG devs). It's still at oi_148, so perhaps just an upgrade to 151 might help? Or will I need to patch the system with later code?
As oi_148 was released before May 2011, it stands to reason you don't
have the fix. As George has been good enough to give the fix ID, you
should be able to see whether the change was committed in 151 by looking
through its log:

http://hg.openindiana.org/sustaining/oi_151a/illumos-gate/shortlog



-------------------------------------------
illumos-developer
Archives: https://www.listbox.com/member/archive/182179/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182179/21175072-c42b328b
Modify Your Subscription: https://www.listbox.com/member/?member_id=21175072&id_secret=21175072-605db409
Powered by Listbox: http://www.listbox.com
Roy Sigurd Karlsbakk
2012-01-20 21:23:47 UTC
Permalink
Post by Bayard G. Bell
Post by Roy Sigurd Karlsbakk
What problem are you having in particular? My colleague George Wilson has done some
work to improve performance on imbalanced vdevs, which may be sufficient for you.
Be sure you are running a build with the fix for 1051 , integrated in May 2011.
The issue we're seeing is rather terrible performance, especially
during scrub, on this system. It has sufficient amounts of L2ARC
(10x256GB) and SLOG (a 100GB mirror) (see attached zpool iostat -v -
the mirror with c14t21d0 and c14t22d0 are the SLOG devs). It's still
at oi_148, so perhaps just an upgrade to 151 might help? Or will I
need to patch the system with later code?
As oi_148 was released before May 2011, it stands to reason you don't
have the fix. As George has been good enough to give the fix ID, you
should be able to see whether the change was committed in 151 by looking
http://hg.openindiana.org/sustaining/oi_151a/illumos-gate/shortlog
Thanks, but do I read this correctly when I fail to find #1050 in there at all?

Vennlige hilsener / Best regards

roy
--
Roy Sigurd Karlsbakk
(+47) 97542685
***@karlsbakk.net
http://blogg.karlsbakk.net/
--
I all pedagogikk er det essensielt at pensum presenteres intelligibelt. Det er et elementært imperativ for alle pedagoger å unngå eksessiv anvendelse av idiomer med fremmed opprinnelse. I de fleste tilfeller eksisterer adekvate og relevante synonymer på norsk.
Bayard G. Bell
2012-01-20 21:46:34 UTC
Permalink
Post by Roy Sigurd Karlsbakk
Post by Bayard G. Bell
Be sure you are running a build with the fix for 1051 , integrated in May 2011.
http://hg.openindiana.org/sustaining/oi_151a/illumos-gate/shortlog
Thanks, but do I read this correctly when I fail to find #1050 in there at all?
If you look under the correct ID (1051), you'll find it.
Roy Sigurd Karlsbakk
2012-01-23 22:07:34 UTC
Permalink
Post by Bayard G. Bell
Post by Roy Sigurd Karlsbakk
Post by Bayard G. Bell
Be sure you are running a build with the fix for 1051 ,
integrated
in May 2011.
http://hg.openindiana.org/sustaining/oi_151a/illumos-gate/shortlog
Thanks, but do I read this correctly when I fail to find #1050 in there at all?
If you look under the correct ID (1051), you'll find it.
FYI: After upgrading to 151a, the server is again useful, which is nice, buying me time to get a new box setup...

Vennlige hilsener / Best regards

roy
--
Roy Sigurd Karlsbakk
(+47) 97542685
***@karlsbakk.net
http://blogg.karlsbakk.net/
--
I all pedagogikk er det essensielt at pensum presenteres intelligibelt. Det er et elementært imperativ for alle pedagoger å unngå eksessiv anvendelse av idiomer med fremmed opprinnelse. I de fleste tilfeller eksisterer adekvate og relevante synonymer på norsk.
Continue reading on narkive:
Loading...