Dear pgsql community,
I've been using postgres for a long time. Recently I'm doing table
sharding over a bunch of pgsql instances. I'm using multiple tablespaces
one per disk to utilize all the IO bandwidth. Things went on pretty
well, however there is a troublesome problem I have when adding
There has been a previous discussion about this topic, including an attempted
fix by Amit Kapila:
http://www.postgresql.org/message-id/flat/CAA4eK1L=tHmmHDK_KW_ja1_dusJxJF+SGQHi=aps4mdnpk7...@mail.gmail.com/
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to
On Fri, Nov 25, 2016 at 4:02 PM, Ideriha, Takeshi
wrote:
> I applied your fixed patch and new one, and confirmed the applied source
> passed the tests successfully. And I also checked manually the error messages
> were emitted successfully when cr/lf are included
Hi,
while doing some benchmarking, I've once again got confused by the
default settings for checkpoint_flush_after and bgwriter_flush_after.
The sample config says this:
#checkpoint_flush_after = 0 # 0 disables,
# default is 256kB on linux, 0 otherwise
> > [Summary]
> > 1. apply patch and make world
> > -> failed because was mistakenly coded .
> >
> > 2.correct this mistake and make check-world
> > -> got 1 failed test: "'pg_dumpall with \n\r in database name'"
> > because test script cannot createdb "foo\n\rbar"
>
> The attached
2016-11-25 7:44 GMT+01:00 Pavel Stehule :
>
>
> 2016-11-25 3:31 GMT+01:00 Alvaro Herrera :
>
>> Michael Paquier wrote:
>>
>> > Nit: I did not look at the patch in details,
>> > but I find the size of the latest version sent, 167kB, scary as it
>>
2016-11-25 3:31 GMT+01:00 Alvaro Herrera :
> Michael Paquier wrote:
>
> > Nit: I did not look at the patch in details,
> > but I find the size of the latest version sent, 167kB, scary as it
> > complicates review and increases the likeliness of bugs.
>
> Here's the stat.
On 11/25/2016 07:19 AM, Tsunakawa, Takayuki wrote:
Specifying multiple hosts is a new feature to be introduced in v10, so that's
here:
https://www.postgresql.org/docs/devel/static/libpq-connect.html
Thanks, I had missed that patch. If we add support for multiple hosts I
think we should also
From: pgsql-hackers-ow...@postgresql.org
> sense to add support for multiple hostaddrs. For consitency's sake if
> nothing else.
Yes, consistency and performance. The purpose of hostaddr is to speed up
connection by eliminating DNS lookup, isn't it? Then, some users should want
to specify
Craig Ringer writes:
> PGDLLIMPORT is free, so the question should be "is there a reason not
> to add it here?".
TBH, my basic complaint about it is that I do not like Microsoft's tool
chain assuming that it's entitled to demand that people sprinkle
Microsoft-specific
Hello,
At Fri, 25 Nov 2016 06:51:43 +0100, Pavel Stehule
wrote in
> I am sure about benefit of all patches - but it is lot of changes in one
> moment, and it is not necessary in this moment.
>
>
On Fri, Nov 25, 2016 at 10:41 AM, Tsunakawa, Takayuki <
tsunakawa.ta...@jp.fujitsu.com> wrote:
> I agree that pg_conn_host should have hostaddr in addition to host, and
PQhost() return host when host is specified with/without hostaddr specified.
typedef struct pg_conn_host
+{
*+ char *host; /*
2016-11-25 1:44 GMT+01:00 Robert Haas :
> On Thu, Nov 24, 2016 at 6:20 PM, Pavel Stehule
> wrote:
> >> I think that the whole emphasis on whether and to what degree this is
> >> like Oracle is somewhat misplaced. I would look at it a different
>
On 11/25/2016 06:11 AM, Tsunakawa, Takayuki wrote:
However, I wonder whether the hostaddr parameter should also accept multiple IP
addresses.
Yeah, I too thought about if we should fix that. I feel like it would
make sense to add support for multiple hostaddrs. For consitency's sake
if
2016-11-25 2:24 GMT+01:00 Kyotaro HORIGUCHI :
> Hello,
>
> Thank you for looking this long-and-bothersome patch.
>
>
> At Wed, 23 Nov 2016 07:12:00 +0100, Pavel Stehule
> wrote in vd9fgpsgv...@mail.gmail.com>
> > Hi
> >
> > 2016-11-15
From: pgsql-hackers-ow...@postgresql.org
> [mailto:pgsql-hackers-ow...@postgresql.org] On Behalf Of Andreas Karlsson
> On 11/24/2016 10:38 PM, Andreas Karlsson wrote:
> > To me it feels like the proper fix would be to make PQHost() return
> > the value of the host parameter rather than the
Amit Kapila writes:
> On Fri, Nov 25, 2016 at 3:26 AM, Andreas Seltenreich
> wrote:
>> just caught another InitPlan below Gather with the recent patches in
>> (master as of 4cc6a3f). Recipe below.
> I think this problem exists since commit
>
>
> I assume you meant "...right after the column name"?
>
> I will modify the grammar to allow that way then, so that the following
> will work:
>
> create table p1 partition of p (
> a primary key
> ) for values in (1);
>
That seems to be non-intuitive as well. The way it's written it
On Fri, Nov 25, 2016 at 3:26 AM, Andreas Seltenreich wrote:
> Hi,
>
> just caught another InitPlan below Gather with the recent patches in
> (master as of 4cc6a3f). Recipe below.
>
I think this problem exists since commit
110a6dbdebebac9401b43a8fc223e6ec43cd4d10 where we
On Fri, Nov 25, 2016 at 8:15 AM, Andreas Karlsson wrote:
> On 11/24/2016 10:38 PM, Andreas Karlsson wrote:
>> To me it feels like the proper fix would be to make PQHost() return the
>> value of the host parameter rather than the hostaddr (maybe add a new
>> field in the
Amit Langote wrote:
> On 2016/11/25 4:36, Alvaro Herrera wrote:
> > I think CREATE TABLE OF is pretty much a corner case. I agree that
> > allowing the constraint right after the constraint name is more
> > intuitive.
>
> I assume you meant "...right after the column name"?
Eh, right.
--
Hi
Here is a tiny patch to fix a typo in execParallel.c.
--
Thomas Munro
http://www.enterprisedb.com
typo.patch
Description: Binary data
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
On Thu, Nov 24, 2016 at 6:13 AM, Amit Langote
wrote:
> On 2016/11/23 4:50, Robert Haas wrote:
>> On Tue, Nov 22, 2016 at 4:15 AM, Amit Langote
>> wrote:
The easiest thing to do might be to just enforce that all of the
Michael Paquier wrote:
> Nit: I did not look at the patch in details,
> but I find the size of the latest version sent, 167kB, scary as it
> complicates review and increases the likeliness of bugs.
Here's the stat. Note that removing the functionality as discussed
would remove all of
On Wed, Nov 23, 2016 at 12:53 PM, Tom Lane wrote:
> Haribabu Kommi writes:
> > On Wed, Nov 23, 2016 at 1:42 AM, Tom Lane wrote:
> >> The precedent of int4/int8/float4/float8 is that SQL data types should
> >> be named after
On Wed, Nov 23, 2016 at 4:19 PM, Thomas Munro
wrote:
> On Wed, Oct 12, 2016 at 4:19 PM, Thomas Munro
> wrote:
>> Here it is, now split into two parts: one patch
>> to add an amskip operation, and another to consider using it to
>>
Hello,
Thank you for looking this long-and-bothersome patch.
At Wed, 23 Nov 2016 07:12:00 +0100, Pavel Stehule
wrote in
> Hi
>
> 2016-11-15 12:26 GMT+01:00 Kyotaro HORIGUCHI
On 25 November 2016 at 02:44, Alvaro Herrera wrote:
> Craig Ringer wrote:
>
>> Updated to correct the other expected file, since there's an alternate.
>
> FWIW I don't know what you did here, but you did not patch the
> alternate expected file.
Damn. Attached the first
On 25 November 2016 at 02:47, Alvaro Herrera wrote:
> Craig Ringer wrote:
>> On 27 October 2016 at 00:42, Robert Haas wrote:
>> > On Wed, Oct 26, 2016 at 7:17 AM, Andres Freund wrote:
>> >> On 2016-09-23 16:04:32 -0400, Tom
On 2016/11/25 4:36, Alvaro Herrera wrote:
> Amit Langote wrote:
>> On 2016/11/24 15:10, Ashutosh Bapat wrote:
>>> On Thu, Nov 24, 2016 at 11:34 AM, Amit Langote wrote:
>
You have to specify column constraints using the keywords WITH OPTIONS,
like below:
create table p1
On 25 November 2016 at 07:36, Michael Paquier wrote:
> On Thu, Nov 24, 2016 at 11:01 PM, Magnus Hagander wrote:
>> On Thu, Nov 24, 2016 at 2:58 PM, Craig Ringer wrote:
>> My guess is that PGDLLIMPORT has been added
On Fri, Nov 25, 2016 at 3:31 AM, Pavel Stehule wrote:
> 2016-11-24 18:51 GMT+01:00 Tom Lane :
>> contrib/xml2 has always relied on libxslt for xpath functionality.
>> Can we do that here instead of writing, debugging, and documenting
>> a pile of new
On Thu, Nov 24, 2016 at 6:20 PM, Pavel Stehule wrote:
>> I think that the whole emphasis on whether and to what degree this is
>> like Oracle is somewhat misplaced. I would look at it a different
>> way. We've talked many times over the years about how PostgreSQL is
>>
On Thu, Nov 24, 2016 at 11:01 PM, Magnus Hagander wrote:
> On Thu, Nov 24, 2016 at 2:58 PM, Craig Ringer wrote:
> My guess is that PGDLLIMPORT has been added explicitly when somebody needed
> it for something, without any actual thought. I can't say I
> I think that the whole emphasis on whether and to what degree this is
> like Oracle is somewhat misplaced. I would look at it a different
> way. We've talked many times over the years about how PostgreSQL is
> optimized for aborts. Everybody that I've heard comment on that issue
> thinks that
On 11/24/2016 10:38 PM, Andreas Karlsson wrote:
To me it feels like the proper fix would be to make PQHost() return the
value of the host parameter rather than the hostaddr (maybe add a new
field in the pg_conn_host struct). But would be a behaviour change which
might break someones application.
On Thu, Nov 24, 2016 at 5:29 AM, David Rowley
wrote:
> There's 11 functions which are marked immutable, but are marked as
> parallel unsafe.
>
> postgres=# select proname from pg_proc where provolatile = 'i' and
> proparallel = 'u';
>proname
>
On Thu, Nov 24, 2016 at 11:06 AM, Greg Stark wrote:
> Fwiw, Oracle does not use the undo log for snapshot fetches. It's used
> only for transaction rollback and recovery.
>
> For snapshot isolation Oracle has yet a *third* copy of the data in a
> space called the "rollback
I propose to rename allow_long to huge_ok. "Huge" is the terminology
used by palloc anyway. I'd keep makeLongStringInfo() and
initLongStringInfo() though as interface, because using Huge instead of
Long there looks strange. Not wedded to that, though (particularly as
it's a bit inconsistent).
Hi,
just caught another InitPlan below Gather with the recent patches in
(master as of 4cc6a3f). Recipe below.
regards,
andreas
set max_parallel_workers_per_gather = 2;
set min_parallel_relation_size = 0;
set parallel_setup_cost = 0;
set parallel_tuple_cost = 0;
explain select 1 from
Hi,
The SSL test suite (src/test/ssl) is broken in the master since commit
9a1d0af4ad2cbd419115b453d811c141b80d872b, which is Robert's refactoring
of getting the server hostname for GSS, SSPI, and SSL in libpq.
The error we get in the test suite:
# Running: psql -X -A -t -c SELECT
Daniel Verite wrote:
> Here's an updated patch. Compared to the previous version:
>
> - removed CopyStartSend (per comment #1 in review)
>
> - renamed flag to allow_long (comment #2)
>
> - resetStringInfo no longer resets the flag (comment #3).
>
> - allowLongStringInfo() is removed (comment
Amit Langote wrote:
> On 2016/11/24 15:10, Ashutosh Bapat wrote:
> > On Thu, Nov 24, 2016 at 11:34 AM, Amit Langote wrote:
> >> You have to specify column constraints using the keywords WITH OPTIONS,
> >> like below:
> >>
> >> create table p1 partition of p (
> >> a with options primary key
>
Craig Ringer wrote:
> On 27 October 2016 at 00:42, Robert Haas wrote:
> > On Wed, Oct 26, 2016 at 7:17 AM, Andres Freund wrote:
> >> On 2016-09-23 16:04:32 -0400, Tom Lane wrote:
> >>> Looking back over the thread, I see that you also proposed
Craig Ringer wrote:
> Updated to correct the other expected file, since there's an alternate.
FWIW I don't know what you did here, but you did not patch the
alternate expected file.
--
Álvaro Herrerahttps://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA,
Tom Lane wrote:
> Alvaro Herrera writes:
> > I considered the argument here for a bit and I think Craig is right --
>
> FWIW, I agree. We shouldn't require every call site to special-case this,
> and we definitely don't want it to require special cases in SQL code.
>
2016-11-24 18:51 GMT+01:00 Tom Lane :
> Alvaro Herrera writes:
> > Pavel Stehule wrote:
> >> 2016-11-24 0:13 GMT+01:00 Alvaro Herrera :
> >>> Oh my, I just noticed we have a new xpath preprocessor in this patch
> >>> too.
Alvaro Herrera writes:
> Pavel Stehule wrote:
>> 2016-11-24 0:13 GMT+01:00 Alvaro Herrera :
>>> Oh my, I just noticed we have a new xpath preprocessor in this patch
>>> too. Where did this code come from -- did you write it all from
>>>
Pavel Stehule wrote:
> can me send your last work?
Sure, it's in the archives --
https://www.postgresql.org/message-id/20161123233130.oqf7jl6czehy5fiw@alvherre.pgsql
--
Álvaro Herrerahttps://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training &
2016-11-24 18:29 GMT+01:00 Alvaro Herrera :
> Pavel Stehule wrote:
> > Hi
> >
> > 2016-11-24 0:13 GMT+01:00 Alvaro Herrera :
> >
> > > Oh my, I just noticed we have a new xpath preprocessor in this patch
> > > too. Where did this code come from
Pavel Stehule wrote:
> Hi
>
> 2016-11-24 0:13 GMT+01:00 Alvaro Herrera :
>
> > Oh my, I just noticed we have a new xpath preprocessor in this patch
> > too. Where did this code come from -- did you write it all from
> > scratch?
>
> I wrote it from scratch - libxml2
On Thu, Nov 24, 2016 at 2:32 AM, Tsunakawa, Takayuki
wrote:
> IMHO, overall, there should be pros and cons of the current approach and the
> new UNDo one (like Oracle?), depending on the workload. Under update-heavy
> workload, the UNDO method may be better.
> FWIW, while this is basically true, the idea of repurposing UNDO to be
> usable for MVCC is definitely an Oracleism. Mohan's ARIES paper says
> nothing about MVCC.
> For snapshot isolation Oracle has yet a *third* copy of the data in a
> space called the "rollback segment(s)".
UNDO and
Alvaro Herrera writes:
> I considered the argument here for a bit and I think Craig is right --
FWIW, I agree. We shouldn't require every call site to special-case this,
and we definitely don't want it to require special cases in SQL code.
(And I'm for back-patching,
On Thu, Nov 24, 2016 at 04:06:14PM +, Greg Stark wrote:
> For snapshot isolation Oracle has yet a *third* copy of the data in a
> space called the "rollback segment(s)". When you update a row in a
> block you save the whole block in the rollback segment. When you try
> to access a block you
On Wed, Nov 23, 2016 at 5:18 PM, Thomas Munro
wrote:
> On Wed, Nov 23, 2016 at 6:01 PM, Peter Geoghegan wrote:
>> * Our behavior with many duplicates in secondary indexes is pretty bad
>> in general, I suspect.
>
> From the pie-in-the-sky
On 23 November 2016 at 04:28, Peter Geoghegan wrote:
> On Tue, Nov 22, 2016 at 7:01 PM, Robert Haas wrote:
>> This basic DO-UNDO-REDO protocol has been well-understood for
>> decades.
>
> FWIW, while this is basically true, the idea of repurposing UNDO to
I have one more question,
In V1 we were calling dsa_detach in ExecParallelCleanup and in
ParallelQueryMain, but it's removed in v2.
Any specific reason ?
Does this need to be used differently ?
ExecParallelCleanup(ParallelExecutorInfo *pei)
{
+ if (pei->area != NULL)
+ {
+
On Wed, Nov 23, 2016 at 5:42 PM, Thomas Munro
wrote:
> ... or we could allow DSA areas to be constructed inside existing
> shmem, as in the attached patch which requires dsa_create_in_place,
> from the patch at
>
On Thu, 24 Nov 2016 15:01:33 +0100
Magnus Hagander wrote:
> On Thu, Nov 24, 2016 at 2:58 PM, Craig Ringer
> wrote:
>
> > Hi all
> >
> > Noticed this while reading something unrelated
> >
> > extern PGDLLIMPORT pid_t PostmasterPid;
> > extern bool
On Thu, Nov 24, 2016 at 10:13:30AM +0100, Magnus Hagander wrote:
> On Thu, Nov 24, 2016 at 2:26 AM, Bruce Momjian wrote:
>
> On Mon, Nov 14, 2016 at 08:43:12PM +, Greg Stark wrote:
> > That said, I don't think the "maintain clustering a bit better using
> >
On Wed, Nov 23, 2016 at 11:35:38PM -0800, Peter Geoghegan wrote:
> On Wed, Nov 23, 2016 at 11:32 PM, Tsunakawa, Takayuki
> wrote:
> > IMHO, overall, there should be pros and cons of the current approach and
> > the new UNDo one (like Oracle?), depending on the
On 11/24/2016 02:49 PM, Andreas Karlsson wrote:
Thanks for finding this. I will look at this more once I get home, but
the tests do not fail on my computer. I wonder what I do differently.
What versions of Perl and OpenSSL do you run and how did you run the
tests when the failed? I ran the
On Thu, Nov 24, 2016 at 2:58 PM, Craig Ringer wrote:
> Hi all
>
> Noticed this while reading something unrelated
>
> extern PGDLLIMPORT pid_t PostmasterPid;
> extern bool IsPostmasterEnvironment;
> extern PGDLLIMPORT bool IsUnderPostmaster;
> extern bool
On Thu, Nov 24, 2016 at 2:58 PM, Craig Ringer wrote:
> Hi all
>
> Noticed this while reading something unrelated
>
> extern PGDLLIMPORT pid_t PostmasterPid;
> extern bool IsPostmasterEnvironment;
> extern PGDLLIMPORT bool IsUnderPostmaster;
> extern bool
Hi all
Noticed this while reading something unrelated
extern PGDLLIMPORT pid_t PostmasterPid;
extern bool IsPostmasterEnvironment;
extern PGDLLIMPORT bool IsUnderPostmaster;
extern bool IsBackgroundWorker;
extern PGDLLIMPORT bool IsBinaryUpgrade;
I don't see any sane reason for some of those to
On 11/24/2016 08:46 AM, Michael Paquier wrote:
On Sat, Nov 12, 2016 at 3:42 AM, Andreas Karlsson wrote:
On 11/11/2016 07:40 PM, Andreas Karlsson wrote:
Here is a new version of the patch with the only differences;
1) The SSL tests have been changed to use reload rather
I considered the argument here for a bit and I think Craig is right --
FrozenXid eventually makes it to a tuple's xmin where it becomes a burden
to the caller, making our interface bug-prone -- sure you can
special-case it, but you don't until it first happens ... and it may not
until you're deep
Local server log has the line and remote table log is empty (it is
configured for minimum warning and when I produce one it appears in log OK)
And I have new details - it happens on some additional environments - not
constantly. Some hours it happens every time, then just stops appearing:
On Wed, 23 Nov 2016 23:08:18 -0600
"Karl O. Pinc" wrote:
> On Wed, 23 Nov 2016 03:21:05 -0600
> "Karl O. Pinc" wrote:
>
> > On Sat, 19 Nov 2016 12:58:47 +0100
> > Gilles Darold wrote:
> >
> > > ... attached v14 of the patch.
>
>
On 2016/11/22 18:28, Ashutosh Bapat wrote:
The comments should explain why is the assertion true.
+/* Shouldn't be NIL */
+Assert(tlist != NIL);
I noticed that I was wrong; in the Assertion the tlist can be empty. An
example for such a case is:
SELECT 1 FROM (SELECT c1 FROM
On Wed, Nov 23, 2016 at 10:19 PM, Catalin Iacob
wrote:
On Tue, Nov 22, 2016 at 8:38 AM, Tsunakawa, Takayuki <
tsunakawa.ta...@jp.fujitsu.com> wrote:
>> If you want to connect to a server where the transaction is read-only,
then shouldn't the connection parameter be
On Wed, Nov 16, 2016 at 3:10 PM, Rushabh Lathia
wrote:
>
>
> On Mon, Nov 14, 2016 at 3:51 PM, Thomas Munro <
> thomas.mu...@enterprisedb.com> wrote:
>
>> On Sat, Nov 12, 2016 at 1:56 AM, Rushabh Lathia
>> wrote:
>> > On Fri, Nov 4, 2016 at
On 2016/11/23 4:50, Robert Haas wrote:
> On Tue, Nov 22, 2016 at 4:15 AM, Amit Langote
> wrote:
>>> The easiest thing to do might be to just enforce that all of the
>>> partition key columns have to be not-null when the range-partitioned
>>> table is defined, and
On 23 November 2016 at 21:19, Thomas Munro
wrote:
> Worth pursuing? Does amskip suck? Does anyone have better ideas,
> either for how to do the low level skip or the higher level Index Skip
> Scan, or perhaps a completely different way of looking at this?
I have
There's 11 functions which are marked immutable, but are marked as
parallel unsafe.
postgres=# select proname from pg_proc where provolatile = 'i' and
proparallel = 'u';
proname
-
_pg_expandarray
_pg_keysequal
_pg_truetypid
_pg_truetypmod
On Mon, Nov 21, 2016 at 1:59 PM, Kouhei Kaigai wrote:
> Hello,
>
> The attached patch is a revised version of pass-down LIMIT to FDW/CSP.
>
> Below is the updates from the last version.
>
> 'ps_numTuples' of PlanState was declared as uint64, instead of long
> to avoid
On 2016/11/24 18:20, Ashutosh Bapat wrote:
I wrote:
You missed the point; the foreignrel->reltarget->exprs doesn't contain
any
PHVs, so the tlist created by build_tlist_to_depase will be guaranteed to
be
one-to-one with the foreignrel->reltarget->exprs.
You wrote:
It's guaranteed now, but
>
>
build_tlist_to_depase() calls pull_var_nodes() before creating the
tlist, whereas the code that searches does not do that. Code-wise
those are not the same things.
>
>
>>> You missed the point; the foreignrel->reltarget->exprs doesn't contain
>>> any
>>> PHVs, so the tlist
On Thu, Nov 24, 2016 at 2:26 AM, Bruce Momjian wrote:
> On Mon, Nov 14, 2016 at 08:43:12PM +, Greg Stark wrote:
> > That said, I don't think the "maintain clustering a bit better using
> > BRIN" is a bad idea. It's just the bit about turning a table
> > append-only to deal
On 2016/11/24 17:39, Ashutosh Bapat wrote:
On Thu, Nov 24, 2016 at 1:27 PM, Etsuro Fujita
wrote:
On 2016/11/24 16:46, Ashutosh Bapat wrote:
table will be misleading as subquery can represent a join and
corresponding alias would represent the join. Relation is
On Thu, Nov 24, 2016 at 1:27 PM, Etsuro Fujita
wrote:
> On 2016/11/24 16:46, Ashutosh Bapat wrote:
Sorry. I think the current version is better than previous one. The
term "subselect alias" is confusing in the previous version. In the
current
On 2016/11/24 16:46, Ashutosh Bapat wrote:
Sorry. I think the current version is better than previous one. The
term "subselect alias" is confusing in the previous version. In the
current version, "Get the relation and column alias for a given Var
node," we need to add word "identifiers" like
83 matches
Mail list logo