On Mon, Jan 18, 2010 at 2:29 AM, Greg Stark gsst...@mit.edu wrote:
Bloom filters need to be sized to have n bits per set element to
maintain a targeted false positive rate. And that false positive rate
would be higher than just having an n bit hash for each set element.
Normally they have the
On Mon, 2010-01-18 at 08:28 +0200, Heikki Linnakangas wrote:
Simon Riggs wrote:
On Sun, 2010-01-17 at 22:57 +0200, Heikki Linnakangas wrote:
It would be good if there was a keepalive WAL record with a
timestamp on it generated every N seconds while in streaming mode.
Yeah, that would
On Fri, 2010-01-15 at 13:28 +0200, Heikki Linnakangas wrote:
I think it's a premature optimization to skip writing the records if
we've written in the same session already. Especially with the
'reason'
information added to the records, it's nice to have a record of each
such operation. All
Joshua D. Drake j...@commandprompt.com wrote:
My question is, do we have any interest in working on getting this into
core?
We had a discussion how replication projects work together with the core
in the developer meeting on PGCon 2009 Japan.
This patch no longer applies. Could you rebase it?
Done (I think). Added a couple of simple tests for bit overlay.
I didn't include the catversion.h changes: obviously the CATALOG_VERSION_NO has
to be changed.
Leonardo
Index: src/backend/utils/adt/varbit.c
On Mon, Jan 18, 2010 at 5:22 AM, Heikki Linnakangas
heikki.linnakan...@enterprisedb.com wrote:
This could be because the win32 socket emulation layer simply wasn't
designed to deal with non-blocking sockets. Specifically, it actually
*always* sets the socket to non-blocking mode, and then uses
On Mon, 2010-01-18 at 08:28 +0200, Heikki Linnakangas wrote:
Do we need a new record type for that, is there a handy record type to
bounce from?
After starting streaming, slices of WAL are sent as CopyData messages.
The CopyData payload begins with an XLogRecPtr, followed by the WAL
On Mon, Jan 18, 2010 at 5:45 PM, Simon Riggs si...@2ndquadrant.com wrote:
How accurate is this now? With regard to remaining items of work.
http://wiki.postgresql.org/wiki/Streaming_Replication
Not accurate. I'll correct that and provide the link from
v8.5 Open Items page to that.
Regards,
--
On Mon, Jan 18, 2010 at 10:30, Fujii Masao masao.fu...@gmail.com wrote:
On Mon, Jan 18, 2010 at 5:22 AM, Heikki Linnakangas
heikki.linnakan...@enterprisedb.com wrote:
This could be because the win32 socket emulation layer simply wasn't
designed to deal with non-blocking sockets. Specifically,
On Mon, Jan 18, 2010 at 00:52, Greg Smith g...@2ndquadrant.com wrote:
Magnus Hagander wrote:
Maybe this should be Unrecognized reset target: %s, target, and also
a errhint() saying which targets are allowed. Thoughts?
That seems reasonable. The other thing I realized is that I forgot to
(2010/01/10 22:25), Magnus Hagander wrote:
The attached patch implements RADIUS authentication (RFC2865-compatible).
The main usecase for me in this is the ability to use (token based)
one-time-password systems easily with PostgreSQL. These systems almost
always support RADIUS, and the
On Mon, Jan 18, 2010 at 6:40 PM, Magnus Hagander mag...@hagander.net wrote:
SSL_read calls into pqwin32_recv(), so you have the same problem. (see
my_sock_read() and my_sock_write() in be-secure.c)
Oh, I confirmed that. Thanks!
Can we prevent SSL_read from being blocked in the renegotiation
Hi,
we have found that auto-prepare causes a problem if the connection
is closed and re-opened and the previously prepared query is issued
again. The application gets back a error code -201 which seems bogus
and it turned out to be a missing
return (false);
after
On Mon, Jan 18, 2010 at 6:35 PM, Fujii Masao masao.fu...@gmail.com wrote:
On Mon, Jan 18, 2010 at 5:45 PM, Simon Riggs si...@2ndquadrant.com wrote:
How accurate is this now? With regard to remaining items of work.
http://wiki.postgresql.org/wiki/Streaming_Replication
Not accurate. I'll
2010/1/17 Tom Lane t...@sss.pgh.pa.us:
I want to do something about the open item discussed in this thread:
http://archives.postgresql.org/message-id/2009081446.ga25...@depesz.com
The right way to handle that, IMO, is to create pg_constraint rows for
triggers created via CREATE CONSTRAINT
On Mon, 2010-01-18 at 20:18 +0900, Fujii Masao wrote:
On Mon, Jan 18, 2010 at 6:35 PM, Fujii Masao masao.fu...@gmail.com wrote:
On Mon, Jan 18, 2010 at 5:45 PM, Simon Riggs si...@2ndquadrant.com wrote:
How accurate is this now? With regard to remaining items of work.
Fujii Masao wrote:
On Mon, Jan 18, 2010 at 6:35 PM, Fujii Masao masao.fu...@gmail.com wrote:
On Mon, Jan 18, 2010 at 5:45 PM, Simon Riggs si...@2ndquadrant.com wrote:
How accurate is this now? With regard to remaining items of work.
http://wiki.postgresql.org/wiki/Streaming_Replication
Not
Yeah, (1) rand isn't a good random number generator and (2) fooling with
I tested rand() and random() generator and results was close to the same, and
rand() was even slightly better.
the main random number sequence is user-unfriendly. If you need a
That's really easy to fix.
private
So the other thread on bloom filter indexes got a discussion going in
real space about other neat things you can do with them and there were
two that seemed useful for Postgres.
We could efficiently get an fairly accurate estimate of the number of
distinct values when doing analyze if we scan the
Simon Riggs wrote:
What were the blockers that prevented sync rep from being included? I
must have missed the discussion on that part.
For one, figuring out how to send back the notifications about WAL
applied in standby, and all the IPC required for that.
Streaming replication is a complex
Simon Riggs wrote:
Can we call that XLogReportUnloggedStatement() or similar?
XlogSkipLogging() sounds like a request rather than a mark/report/record
type of action.
Agreed. I vote for XLogReportUnloggedOperation(). I'll change it to that
before committing, unless Fujii beats me to it.
--
So for example if your bloom filter is 4 bits per column your error
rate for a single column where clause will be 1/2^(4/1.44) or a little
worse than returning 1 record in 7. If you test two or three columns
then it'll be about 1 in 49 or 1 in 343.
Hmm, I don't understand your calculations. In
2010/1/18 Teodor Sigaev teo...@sigaev.ru:
So for example if your bloom filter is 4 bits per column your error
rate for a single column where clause will be 1/2^(4/1.44) or a little
worse than returning 1 record in 7. If you test two or three columns
then it'll be about 1 in 49 or 1 in 343.
Another idea of how to make use of bloom filters:
It should be possible to implement a gist opclass over a bloomfilter
data type which implements the operation int membership in int[].
All you need is a function which takes an int[] and returns a bloom
filter. Then your union operation is to
On Mon, Jan 18, 2010 at 9:17 PM, Heikki Linnakangas
heikki.linnakan...@enterprisedb.com wrote:
Agreed. I vote for XLogReportUnloggedOperation().
OK.
I'll change it to that
before committing, unless Fujii beats me to it.
Yeah, please go ahead.
Regards,
--
Fujii Masao
NIPPON TELEGRAPH AND
Fujii Masao masao.fu...@gmail.com writes:
When I configured a cascaded standby (i.e, made the additional
standby server connect to the standby), I got the following
errors, and a cascaded standby didn't start replication.
ERROR: timeline 0 of the primary does not match recovery target
Heikki Linnakangas heikki.linnakan...@enterprisedb.com writes:
Simon Riggs wrote:
Do we need a new record type for that, is there a handy record type to
bounce from?
After starting streaming, slices of WAL are sent as CopyData messages.
The CopyData payload begins with an XLogRecPtr,
On Mon, 2010-01-18 at 09:31 -0500, Tom Lane wrote:
Fujii Masao masao.fu...@gmail.com writes:
When I configured a cascaded standby (i.e, made the additional
standby server connect to the standby), I got the following
errors, and a cascaded standby didn't start replication.
ERROR:
2010/1/18 Tom Lane t...@sss.pgh.pa.us:
Magnus Hagander mag...@hagander.net writes:
Which shows one potentially big problem - since we're calling select()
from inside libpq, it's not calling our signal emulation layer
compatible select(). This means that at this point, walreceiver is
not
2010/1/17 Heikki Linnakangas heikki.linnakan...@enterprisedb.com:
Magnus Hagander wrote:
Which shows one potentially big problem - since we're calling select()
from inside libpq, it's not calling our signal emulation layer
compatible select(). This means that at this point, walreceiver is
not
Il 16/01/2010 14:21, Matteo Beccati ha scritto:
Il 16/01/2010 11:48, Dimitri Fontaine ha scritto:
Matteo Beccatip...@beccati.com writes:
Anyway, I've made further changes and I would say that at this point
the PoC
is feature complete. There surely are still some rough edges and a few
things to
Teodor Sigaev teo...@sigaev.ru writes:
Yeah, (1) rand isn't a good random number generator and (2) fooling with
I tested rand() and random() generator and results was close to the same, and
rand() was even slightly better.
That is true on some platforms, but on others rand() is very bad.
2010/1/18 Matteo Beccati p...@beccati.com:
Il 16/01/2010 14:21, Matteo Beccati ha scritto:
Il 16/01/2010 11:48, Dimitri Fontaine ha scritto:
Matteo Beccatip...@beccati.com writes:
Anyway, I've made further changes and I would say that at this point
the PoC
is feature complete. There
Dean Rasheed dean.a.rash...@googlemail.com writes:
Agreed. That's much neater. However, it does introduce a change in
behaviour - if you have 2 constraints with the same name in different
schemas, one deferrable, and one not, and the non-deferrable one is
first on the search path, then you'll
Tom Lane wrote:
Heikki Linnakangas heikki.linnakan...@enterprisedb.com writes:
Simon Riggs wrote:
Do we need a new record type for that, is there a handy record type to
bounce from?
After starting streaming, slices of WAL are sent as CopyData messages.
The CopyData payload begins with an
Magnus Hagander mag...@hagander.net writes:
Also, I tink one of the main issues with the archives today that
people bring up is the inability to have threads cross months. I think
that should be fixed. Basically, get rid of the grouping by month for
a more dynamic way to browse.
Clic a mail
Heikki Linnakangas heikki.linnakan...@enterprisedb.com writes:
Tom Lane wrote:
Whether or not anyone bothers with the timestamp message, I think adding
a message type header is a Must Fix item. A protocol with no provision
for extension is certainly going to bite us in the rear before long.
Magnus Hagander wrote:
2010/1/17 Heikki Linnakangas heikki.linnakan...@enterprisedb.com:
We could replace the blocking PQexec() calls with PQsendQuery(), and use
the emulated version of select() to wait.
Hmm. That would at least theoretically work, but aren't there still
places we may end
Greg Sabino Mullane wrote:
-BEGIN PGP SIGNED MESSAGE-
Hash: RIPEMD160
I mentioned earlier that buildfarm member jaguar (that's the one that
builds with CLOBBER_CACHE_ALWAYS) was showing suspicious intermittent
failures.
Tom, this brings up another question: is there any
Magnus Hagander wrote:
2010/1/18 Matteo Beccati p...@beccati.com:
My question now is... what next? :)
Gee, I disappear for a week and look what happens -- we get streaming
replication, a revamped archives site, and maybe something else that I
haven't seen yet. I love it :-)
If it wasn't
I reviewed this patch today.
Overview:
Almost everything is OK. Applied with few hunks (in psql/describle.c 2
lines offset), compiled without warnings, passed regression tests. The
results of advertised queries are as expected. Coding style is of
course satisfied. Since this is utility changes
Hi,
Quoting Kevin Grittner kevin.gritt...@wicourts.gov:
I strongly encourage you to set that up on git.postgresql.org.
I'm about to provide git repositories for Postgres-R anyway, so I've
setup two projects on git.postgres-r.org:
dtester: that's the driver/harness code
postgres-dtests: a
Looking at this patch for the commitfest I have a few questions.
1) You said you added an fsync of the new directory -- where is that I
don't see it anywhere.
2) Why does the second pass to do the fsyncs read through fromdir to
find all the filenames. I find that odd and counterintuitive. It
Tom Lane wrote:
Speaking of which, just where is the defense that makes sure that
walsender and walreceiver are compatible? We should be checking not
only version, but all of the configuration variables that are embedded
in pg_control.
That happens at startup when pg_control is read, before
Il 18/01/2010 15:55, Magnus Hagander ha scritto:
If it wasn't for the fact that we're knee deep in two other major
projects for the infrastructure team right now, I'd be all over this
:-) But we really need to complete that before we put anything new in
production here.
Sure, that's completely
Il 18/01/2010 16:19, Dimitri Fontaine ha scritto:
Magnus Hagandermag...@hagander.net writes:
Also, I tink one of the main issues with the archives today that
people bring up is the inability to have threads cross months. I think
that should be fixed. Basically, get rid of the grouping by month
On Mon, Jan 18, 2010 at 18:31, Matteo Beccati p...@beccati.com wrote:
Il 18/01/2010 15:55, Magnus Hagander ha scritto:
If it wasn't for the fact that we're knee deep in two other major
projects for the infrastructure team right now, I'd be all over this
:-) But we really need to complete that
On Mon, Jan 18, 2010 at 18:35, Matteo Beccati p...@beccati.com wrote:
Il 18/01/2010 16:19, Dimitri Fontaine ha scritto:
Magnus Hagandermag...@hagander.net writes:
Also, I tink one of the main issues with the archives today that
people bring up is the inability to have threads cross months.
On Sun, Jan 17, 2010 at 2:04 PM, Pavel Stehule pavel.steh...@gmail.com wrote:
I rewrote patch so now interface for PQescapeIdentConn is same as
PQescapeStringConn
@3. I though so the protection under incomplete multibyte chars are
enought - missing bytes are replaced by space - like
2010/1/18 Robert Haas robertmh...@gmail.com:
On Sun, Jan 17, 2010 at 2:04 PM, Pavel Stehule pavel.steh...@gmail.com
wrote:
I rewrote patch so now interface for PQescapeIdentConn is same as
PQescapeStringConn
@3. I though so the protection under incomplete multibyte chars are
enought -
On Mon, Jan 18, 2010 at 1:52 PM, Pavel Stehule pavel.steh...@gmail.com wrote:
2010/1/18 Robert Haas robertmh...@gmail.com:
On Sun, Jan 17, 2010 at 2:04 PM, Pavel Stehule pavel.steh...@gmail.com
wrote:
I rewrote patch so now interface for PQescapeIdentConn is same as
PQescapeStringConn
@3.
2010/1/18 Robert Haas robertmh...@gmail.com:
On Mon, Jan 18, 2010 at 1:52 PM, Pavel Stehule pavel.steh...@gmail.com
wrote:
2010/1/18 Robert Haas robertmh...@gmail.com:
On Sun, Jan 17, 2010 at 2:04 PM, Pavel Stehule pavel.steh...@gmail.com
wrote:
I rewrote patch so now interface for
Hitoshi Harada umi.tan...@gmail.com writes:
2010/1/17 Tom Lane t...@sss.pgh.pa.us:
That's broken, whether it passes regression tests or not. Not
canonicalizing will mean that you fail to recognize equality to
canonicalized pathkeys, and thus for example execute unnecessary
sorts.
So why
On Tuesday 12 January 2010 01:06:22 Takahiro Itagaki wrote:
Psql shows too many parentheses when it prints triggers with WHEN clause.
postgres=# \d t1
Table public.t1
Column | Type | Modifiers
+-+---
c1 | integer |
Triggers:
mytrig AFTER UPDATE
On Mon, Jan 18, 2010 at 2:20 PM, Pavel Stehule pavel.steh...@gmail.com wrote:
2010/1/18 Robert Haas robertmh...@gmail.com:
On Mon, Jan 18, 2010 at 1:52 PM, Pavel Stehule pavel.steh...@gmail.com
wrote:
2010/1/18 Robert Haas robertmh...@gmail.com:
On Sun, Jan 17, 2010 at 2:04 PM, Pavel Stehule
Robert Haas robertmh...@gmail.com writes:
... Also, I prefer an
API where the escaping function does include the quotes, so I've done
it that way in the attached patch.
IMO this function should act as much like PQescapeStringConn as possible.
Random differences like including or not including
On Mon, Jan 18, 2010 at 3:26 PM, Tom Lane t...@sss.pgh.pa.us wrote:
Robert Haas robertmh...@gmail.com writes:
... Also, I prefer an
API where the escaping function does include the quotes, so I've done
it that way in the attached patch.
IMO this function should act as much like
On sön, 2009-12-06 at 02:21 -0800, Kurt Harriman wrote:
Which ones does it actually offer any benefit for?
MSVC is one.
I seem to recall that somewhere else it was said that MSVC produces
warnings on unused static inline functions. Am I mistaken?
Also, if this is mainly for the benefit of
On sön, 2010-01-17 at 23:54 -0500, Robert Haas wrote:
I always assumed that it was printing the names as the tests finished,
probably because of the way the output is staggered. If they were
printed when the tests were started, the list would be printed all but
simultaneously.
Well, as long
Leonardo F wrote:
Done (I think). Added a couple of simple tests for bit overlay.
I didn't include the catversion.h changes: obviously the
CATALOG_VERSION_NO has to be changed.
[Following the Reviewing A Patch wiki, more or less]
The patch is in context diff format and applies cleanly.
When you create a composite type that already exists, the error message
you get is about 'relation foo already exists'. While true, this can
be confusing, as you didn't plan to create a relation. Therefore, I
propose that we insert a snippet of code that is already used by the
other forms of
Brad T. Sliger b...@sliger.org wrote:
I tried to apply this patch to the latest version of PostgreSQL in git
(bbfc96e). Some of the patch did not apply. Please find attached the
output from patch. The full path of the ruleutils.c.rej is
src/backend/utils/adt/ruleutils.c.rej
The attached
On 1/18/2010 1:20 PM, Peter Eisentraut wrote:
I seem to recall that somewhere else it was said that MSVC produces
warnings on unused static inline functions. Am I mistaken?
MSVC does warn about unused static inline functions. The warning
is prevented by using __forceinline instead of
Kurt Harriman harri...@acm.org writes:
c) Use configure to automate the testing of each build environment
in situ.
The third alternative adapts to new or little-known platforms
with little or no manual intervention.
This argument is bogus unless you can demonstrate a working
2010/1/19 Tom Lane t...@sss.pgh.pa.us:
Hitoshi Harada umi.tan...@gmail.com writes:
In a RANGE offset mode query, for example:
SELECT sum(ten) over (PARTITION BY four ORDER BY four RANGE BETWEEN 2
PRECEDING AND 1 PRECEDING) FROM tenk1
the frame is determined as from the first row which has
On Mon, Jan 18, 2010 at 4:54 PM, Peter Eisentraut pete...@gmx.net wrote:
On sön, 2010-01-17 at 23:54 -0500, Robert Haas wrote:
I always assumed that it was printing the names as the tests finished,
probably because of the way the output is staggered. If they were
printed when the tests were
Hitoshi Harada umi.tan...@gmail.com writes:
2010/1/19 Tom Lane t...@sss.pgh.pa.us:
AFAICS that doesn't mean it can't be the
canonicalized form of the sort key. If a column is dropped out of the
canonical sort key then it's simply redundant, and hence not relevant to
determining the range.
Pavel Stehule pavel.steh...@gmail.com writes:
ok, I accept all comments.
revised version are attached.
Applied with minor editorialization.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
Bernd Helmle wrote:
These are two new functions pg_table_size() and pg_indexes_size().
This patch also changes pg_total_relation_size() to be a shorthand for
pg_table_size() + pg_indexes_size().
Attached is a test program to exercise these new functions. I
thoroughly abuse generate_series
Greg Smith g...@2ndquadrant.com writes:
The only question I'm left with after browsing the patch and staring at
the above results is whether it makes sense to expose a pg_toast_size
function. That would make the set available here capable of handling
almost every situation somebody might
On 1/18/2010 4:42 PM, Tom Lane wrote:
Kurt Harrimanharri...@acm.org writes:
c) Use configure to automate the testing of each build environment
in situ.
The third alternative adapts to new or little-known platforms
with little or no manual intervention.
This argument is bogus
Tom Lane wrote:
I'm inclined to think that table vs. index is the right level of
abstraction for these functions, and that breaking it down further than
that isn't all that helpful. We have the bottom-level information
(per-fork relation size) available for those who really want the
details.
Hi, I'm reviewing your patch and have a couple of comments.
Boszormenyi Zoltan z...@cybertec.at wrote:
we have found that auto-prepare causes a problem if the connection
is closed and re-opened and the previously prepared query is issued
again.
You didn't attach actual test cases for the
On Monday 18 January 2010 16:40:07 Takahiro Itagaki wrote:
Brad T. Sliger b...@sliger.org wrote:
I tried to apply this patch to the latest version of PostgreSQL in git
(bbfc96e). Some of the patch did not apply. Please find attached the
output from patch. The full path of the
Bernd Helmle maili...@oopsware.de writes:
Since i'm not able to finish those other things in time, i wrapped up my
existing code for this issue and came up with the attached patch, which
should implement the behavior Tom proposed. These are two new functions
pg_table_size() and
On Mon, Jan 18, 2010 at 11:42 PM, Simon Riggs si...@2ndquadrant.com wrote:
On Mon, 2010-01-18 at 09:31 -0500, Tom Lane wrote:
Fujii Masao masao.fu...@gmail.com writes:
When I configured a cascaded standby (i.e, made the additional
standby server connect to the standby), I got the following
2010/1/19 Tom Lane t...@sss.pgh.pa.us:
Pavel Stehule pavel.steh...@gmail.com writes:
ok, I accept all comments.
revised version are attached.
Applied with minor editorialization.
thank you
Pavel
regards, tom lane
--
Sent via pgsql-hackers mailing list
Initial comments:
* compiler warnings
ipci.c: In function ‘CreateSharedMemoryAndSemaphores’:
ipci.c:228: warning: implicit declaration of function ‘AsyncShmemInit’
* 2PC
Adds complexity, and I didn't see any clear, easy solution after
reading the thread. I don't see this as a showstopper, so
On Tue, 2010-01-19 at 15:04 +0900, Fujii Masao wrote:
There is an ERROR, but no problem AFAICS. The tli isn't set until end of
recovery because it doesn't need to have been set yet. That shouldn't
prevent retrieving WAL data.
OK. Here is the patch which supports a walsender process
On mån, 2010-01-18 at 16:34 -0800, Kurt Harriman wrote:
MSVC does warn about unused static inline functions. The warning
is prevented by using __forceinline instead of __inline.
Hmm, but forceinline is not the same as inline. Are we confident that
forcing inlining is not going to lead to
80 matches
Mail list logo