On 05/30/2014 12:53 AM, John Lumby wrote:
Date: Thu, 29 May 2014 18:00:28 -0300
Subject: Re: [HACKERS] Extended Prefetching using Asynchronous IO - proposal
and patch
From: klaussfre...@gmail.com
To: hlinnakan...@vmware.com
CC: johnlu...@hotmail.com; pgsql-hackers@postgresql.org
On Thu, May
On Fri, May 30, 2014 at 4:31 AM, Robert Haas robertmh...@gmail.com wrote:
I suspect it would be good, then, to separate this into two patches,
one for back-patch and one not.
Attached are patches for master and 9.3. 9.2 and older versions are
not impacted.
Regards,
--
Michael
diff --git
On 05/30/2014 10:25 AM, Michael Paquier wrote:
On Fri, May 30, 2014 at 4:31 AM, Robert Haas robertmh...@gmail.com wrote:
I suspect it would be good, then, to separate this into two patches,
one for back-patch and one not.
Attached are patches for master and 9.3. 9.2 and older versions are
not
Hi all,
While developing some XML processing queries, i stumbled on an old bug
mentioned in http://wiki.postgresql.org/wiki/Todo#XML: Fix Nested or
repeated xpath() that apparently mess up namespaces.
Source of the bug is that libxml2's xmlNodeDump won't output XML namespace
definitions that
On 2014-05-11 23:30:37 +0200, Andres Freund wrote:
On 2014-05-11 12:24:30 -0400, Tom Lane wrote:
Andres Freund and...@2ndquadrant.com writes:
On 2014-05-10 23:21:34 -0700, Peter Geoghegan wrote:
On Fri, May 9, 2014 at 1:50 PM, Andres Freund and...@2ndquadrant.com
wrote:
And adding
On 12 May 2014 08:15, Andres Freund and...@2ndquadrant.com wrote:
But I concur that in practice, if you're dealing with 16TB tables, it's time
to partition.
Well, we need to improve our partitioning for that to be viable for all
relations. Not having usable foreign and unique keys makes it a
On 2014-05-30 12:00:47 +0100, Simon Riggs wrote:
On 12 May 2014 08:15, Andres Freund and...@2ndquadrant.com wrote:
But I concur that in practice, if you're dealing with 16TB tables, it's
time
to partition.
Well, we need to improve our partitioning for that to be viable for all
On Fri, May 30, 2014 at 12:45 AM, Andrew Dunstan and...@dunslane.net
wrote:
On 05/29/2014 05:41 PM, Josh Kupershmidt wrote:
On Thu, May 29, 2014 at 4:06 PM, Andrew Dunstan and...@dunslane.net
wrote:
Almost all my critters run in VMs (all but jacana and bowerbird).
They're not running on
Hi!
jsonb operators -text, -text,-int, -int use inefficient methods to access
to needed field, proportional O(N/2). Attached patch suggests for text operators
O(log(N)) and for integer - O(1). The fuctions with fast access already are
existed in current code and are used in contains
On 05/28/2014 06:41 PM, Andres Freund wrote:
Hi,
Pushed a fix for it. I am pretty sure it will, but could you still test
that it fixes your problem?
Thanks!
The fix seems to work (I am no longer getting the timeout on slot creation)
Thanks
Andres Freund
--
Sent via pgsql-hackers
On Fri, May 30, 2014 at 4:15 AM, Heikki Linnakangas
hlinnakan...@vmware.com wrote:
We don't want polling... And even if we did, calling aio_suspend() in a
way
that's known to be broken, in a loop, is a pretty crappy way of polling.
Well, as mentioned earlier, it is not broken. Whether
Andres Freund wrote:
On 2014-05-11 23:30:37 +0200, Andres Freund wrote:
Yea. I really don't like to take on such a major project to solve a
minor problem.
What I am thinking about right now is to expose a 'pg_blocknumber'
type. That only does very basic operations and implicitly casts to
The point I'm making is that the scenario your test case exposes is not
an infinite loop of picksplits, but an infinite loop of choose calls.
Thank you, now I see what I missed before. After some brainstorming, it's
possible to use '\0' only as end of string marker. The idea is: when we split
On 05/30/2014 09:35 AM, Teodor Sigaev wrote:
Hi!
jsonb operators -text, -text,-int, -int use inefficient methods
to access to needed field, proportional O(N/2). Attached patch
suggests for text operators O(log(N)) and for integer - O(1). The
fuctions with fast access already are existed in
On 05/29/2014 07:39 PM, Jeff Janes wrote:
It also implicitly tested the xlog parallel write slots thing, as that is
common code to all recovery.
During development, I hit a lot of bugs in that patch by setting
wal_buffers to 32kb (the minimum). Causes more backends to wait for each
other,
On Fri, May 30, 2014 at 3:59 PM, Heikki Linnakangas
hlinnakan...@vmware.com wrote:
If transaction A commits synchronously with commit LSN 1, and transaction B
commits asynchronously with commit LSN 2, B cannot become visible before A.
And we cannot acknowledge B as committed to the client until
On Fri, May 30, 2014 at 06:55:18PM +0400, Teodor Sigaev wrote:
The point I'm making is that the scenario your test case exposes is not
an infinite loop of picksplits, but an infinite loop of choose calls.
Thank you, now I see what I missed before. After some brainstorming,
it's possible to
On 05/30/2014 06:09 PM, Greg Stark wrote:
On Fri, May 30, 2014 at 3:59 PM, Heikki Linnakangas
hlinnakan...@vmware.com wrote:
If transaction A commits synchronously with commit LSN 1, and transaction B
commits asynchronously with commit LSN 2, B cannot become visible before A.
And we cannot
Hi,
On 2014-05-30 17:59:23 +0300, Heikki Linnakangas wrote:
So, here's a first version of the patch. Still very much WIP.
Cool.
One thorny issue came up in discussions with other hackers on this in PGCon:
When a transaction is committed asynchronously, it becomes visible to other
backends
Hello,
(sorry for this long post)
I have patched intarray with 3 additional functions in order to count[distinct]
event IDs
into arrays, whereas the array position correspond to the integer values.
(mimic column oriented storage)
I need to use an array for the event IDs as I don't know how
If we bump the system catalog version, pg_upgrade can mark those indexes
as invalid and output a script to reindex them.
Between 9.3 - 9.4 catalog version is changed anyway.
--
Teodor Sigaev E-mail: teo...@sigaev.ru
WWW:
(reposted, this time with attachment. sorry)
Hello,
(sorry for this long post)
I have patched intarray with 3 additional functions in order to count[distinct]
event IDs
into arrays, whereas the array position correspond to the integer values.
(mimic column oriented storage)
I need to use
On 05/30/2014 11:08 AM, Andrew Dunstan wrote:
On 05/30/2014 09:35 AM, Teodor Sigaev wrote:
Hi!
jsonb operators -text, -text,-int, -int use inefficient methods
to access to needed field, proportional O(N/2). Attached patch
suggests for text operators O(log(N)) and for integer - O(1). The
If we're going to construct varlena objects inside a StringInfo, maybe
we need a proper API for it. Isn't there a danger that data member of
the StringInfo won't be properly aligned to allow us to do this? In any
case, we should get most of the benefit of your patch without this
optimization.
I see that palloc.h says:
The result of palloc() is always word-aligned
void *
repalloc(void *pointer, Size size)
{
...
/*
* Try to detect bogus pointers handed to us, poorly though we can.
* Presumably, a pointer that isn't MAXALIGNED isn't pointing at an
* allocated
The patch really improves access performance to jsonb. On the
delicious bookmarks I got 5 times better performance.Now jsonb
outperforms json on simple access (slide 12 of pgcon presentation) by
103 times !
Oleg
On Fri, May 30, 2014 at 9:35 AM, Teodor Sigaev teo...@sigaev.ru wrote:
Hi!
On 05/30/2014 01:30 PM, Oleg Bartunov wrote:
The patch really improves access performance to jsonb. On the
delicious bookmarks I got 5 times better performance.Now jsonb
outperforms json on simple access (slide 12 of pgcon presentation) by
103 times !
(Oleg, please try not to top-post)
Andrew Dunstan and...@dunslane.net writes:
If we're going to construct varlena objects inside a StringInfo, maybe
we need a proper API for it. Isn't there a danger that data member of
the StringInfo won't be properly aligned to allow us to do this? In any
case, we should get most of the
The question is whether the speedup comes from the reduction in lookup
times or the reduction in string copying. I have a strong suspicion that
it's mostly the first, not the second.
Agree with about Oleg's test, but some users put huge values into
json... Actually, I don't insist, ITSM,
On 05/30/2014 06:27 PM, Andres Freund wrote:
On 2014-05-30 17:59:23 +0300, Heikki Linnakangas wrote:
One thorny issue came up in discussions with other hackers on this in PGCon:
When a transaction is committed asynchronously, it becomes visible to other
backends before the commit WAL record is
(forgot to answer this question)
On 05/30/2014 06:27 PM, Andres Freund wrote:
On 2014-05-30 17:59:23 +0300, Heikki Linnakangas wrote:
* I expanded pg_clog to 64-bits per XID, but people suggested keeping
pg_clog as is, with two bits per commit, and adding a new SLRU for the
commit LSNs beside
Asif Naeem anaeem...@gmail.com writes:
On Fri, May 30, 2014 at 9:49 PM, reiner peterke zedaa...@drizzle.com
wrote:
Since upgrading my mac from os x 10.8 to 10.9, i can no long build
postgres with '--with-pythonâ.
Latest PG 9.4 source code seems building fine with --with-python option on
On 05/30/2014 11:52 AM, Tom Lane wrote:
Asif Naeem anaeem...@gmail.com writes:
On Fri, May 30, 2014 at 9:49 PM, reiner peterke zedaa...@drizzle.com
wrote:
Since upgrading my mac from os x 10.8 to 10.9, i can no long build
postgres with '--with-python’.
Latest PG 9.4 source code seems
On Fri, May 30, 2014 at 08:19:07PM +0400, Teodor Sigaev wrote:
If we bump the system catalog version, pg_upgrade can mark those indexes
as invalid and output a script to reindex them.
Between 9.3 - 9.4 catalog version is changed anyway.
Right. I was thinking of beta users between beta1
Adrian Klaver adrian.kla...@aklaver.com writes:
On 05/30/2014 11:52 AM, Tom Lane wrote:
If you do the ld call by hand without the -syslibroot option, it works.
AFAICS it could never have worked with such an option, so I'm thinking
this is some new misbehavior in the latest version of Xcode.
On Tue, May 27, 2014 at 10:01 PM, Gurjeet Singh gurj...@singh.im wrote:
When the Postgres server is being stopped/shut down, the `Buffer
Saver` scans the
shared-buffers of Postgres, and stores the unique block identifiers of
each cached
block to the disk. This information is saved under the
On Thu, May 29, 2014 at 8:15 PM, Amit Kapila amit.kapil...@gmail.com
wrote:
On Thu, May 29, 2014 at 10:09 PM, Jeff Janes jeff.ja...@gmail.com wrote:
What features in 9.4 need more beta testing for recovery?
Another feature which have interaction with recovery is reduced WAL
for Update
On 05/30/2014 01:31 PM, Tom Lane wrote:
Adrian Klaver adrian.kla...@aklaver.com writes:
On 05/30/2014 11:52 AM, Tom Lane wrote:
If you do the ld call by hand without the -syslibroot option, it works.
AFAICS it could never have worked with such an option, so I'm thinking
this is some new
On Thu, May 29, 2014 at 09:39:56AM -0700, Jeff Janes wrote:
I've applied my partial-write testing harness to several scenarios in 9.4.
So far its found a recovery bug for gin indexes, a recovery bug for btree,
a vacuum bug for btree indexes (with foreign keys, but that is not relevant
to the
On Sat, May 31, 2014 at 3:51 AM, Jeff Janes jeff.ja...@gmail.com wrote:
On Thu, May 29, 2014 at 8:15 PM, Amit Kapila amit.kapil...@gmail.com
wrote:
On Thu, May 29, 2014 at 10:09 PM, Jeff Janes jeff.ja...@gmail.com
wrote:
What features in 9.4 need more beta testing for recovery?
Another
40 matches
Mail list logo
