On Fri, Jan 21, 2011 at 08:44:59AM +0200, Heikki Linnakangas wrote:
> We have enough information in the standby to reconstruct all writes done
> in the master. I gather that's not enough, in order to roll back
> read-only transaction T3 on the standby which would see an anomaly, we'd
> also need
On 21.01.2011 03:19, Dan Ports wrote:
What I'm still not clear on is why that HS is different. Whatever rules
apply on the master must also apply on the standby, immutably. Why is it
we need to pass explicit snapshot information from master to standby? We
don't do that, except at startup for norm
On Fri, Jan 21, 2011 at 1:00 AM, Tom Lane wrote:
> Fujii Masao writes:
>> On Thu, Jan 20, 2011 at 10:53 AM, Tom Lane wrote:
I'm not sure why that's the right solution. Why do you think that we should
not create the tablespace under the $PGDATA directory? I'm not surprised
that peo
On 1/20/2011 7:35 AM, Tom Lane wrote:
Marko Tiikkaja writes:
This seems useful, since the xact lock would be automatically released
if an error happens during "-- do something here" so you wouldn't need
to worry about releasing the lock elsewhere. But I'm not sure this is
safe. Can anyone see
Robert Haas wrote:
> Kevin Grittner wrote:
>> As I mentioned in another email, we might want to throttle this.
>> My thinking was that we could start a timer on capturing a
>> snapshot, and continue to gather new ones as they become
>> available. When you hit the timer limit (maybe 100ms?) you se
> Here is the patch to implement the feature.
>
> 1) pg_terminate_backend() sends SIGUSR1 signal rather than SIGTERM to
>the target backend.
> 2) The infrastructure used for message passing is
>storage/ipc/procsignal.c The new message type for ProcSignalReason
>is "PROCSIG_TERMNINATE_B
> On Fri, Jan 14, 2011 at 1:51 PM, Robert Haas wrote:
>> On Fri, Jan 14, 2011 at 12:29 PM, Simon Riggs wrote:
>>> This whole thing is confused. No change is appropriate here at all.
>>>
>>> We issue ERRCODE_T_R_SERIALIZATION_FAILURE almost all of the time for
>>> recovery conflicts.
>>>
>>> We is
On Thu, Jan 20, 2011 at 9:53 PM, Bernd Helmle wrote:
> I had a look at this for the current CF and the patch looks reasonable to
> me. Some testing shows that the changes are working as intended (at least,
> the wal sender actually receives now signals from SignalSomeChildren() as
> far as the DEB
> Robert Haas wrote:
> Jeff Janes wrote:
>> I found the following message somewhat confusing:
>> ERROR: read-only property must be set before any query
>
> I think what we need here is two messages, this one and a similar one
> that starts with "read-write property...".
>
>> When a subtransacti
Robert Haas wrote:
> Kevin Grittner wrote:
>> Attached is a rebased roll-up of the 3 and 3a patches from last
>> month.
> do you have a link to previous discussion?
http://archives.postgresql.org/pgsql-hackers/2010-12/msg00582.php
That thread seems to break, but if you look at the reference
On Fri, Jan 14, 2011 at 1:51 PM, Robert Haas wrote:
> On Fri, Jan 14, 2011 at 12:29 PM, Simon Riggs wrote:
>> This whole thing is confused. No change is appropriate here at all.
>>
>> We issue ERRCODE_T_R_SERIALIZATION_FAILURE almost all of the time for
>> recovery conflicts.
>>
>> We issue ERRCO
On Sun, Jan 16, 2011 at 6:58 PM, Jeff Janes wrote:
> I found the following message somewhat confusing:
> ERROR: read-only property must be set before any query
I think what we need here is two messages, this one and a similar one
that starts with "read-write property...".
> When a subtransactio
On Mon, Jan 10, 2011 at 11:27 AM, Kevin Grittner
wrote:
> Attached is a rebased roll-up of the 3 and 3a patches from last month.
Sorry to be a dweeb, but do you have a link to previous discussion?
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
--
Se
On Thu, Jan 20, 2011 at 2:22 PM, Noah Misch wrote:
> Okay. I've attached a new patch version based on that strategy.
Thanks. Committed and back-patched to 9.0 (but I didn't use your
regression test).
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
-
On 01/20/2011 09:52 PM, Robert Haas wrote:
On Thu, Jan 20, 2011 at 10:17 AM, XiaoboGu wrote:
Hi,
We are using R to work with 64bit PostgreSQL client libraries, and
to avoid compiler compatibility issues the R development community suggest
using the same compiler for both the main appl
On Thu, Jan 20, 2011 at 10:17 AM, XiaoboGu wrote:
> Hi,
> We are using R to work with 64bit PostgreSQL client libraries, and
> to avoid compiler compatibility issues the R development community suggest
> using the same compiler for both the main application and dlls. So do you
> have any ex
On Thu, Jan 20, 2011 at 9:35 PM, Mark Kirkwood
wrote:
> On 21/01/11 15:24, Robert Haas wrote:
>>
>> On Thu, Jan 20, 2011 at 9:17 PM, Kevin Grittner
>> wrote:
>>>
>>> Right -- God only knows the number of things were filtered out to
>>> leave me with filtered water. What's "filtered" in this cas
On Thu, Jan 20, 2011 at 6:19 PM, Noah Misch wrote:
> On Thu, Jan 20, 2011 at 09:36:11PM +, Simon Riggs wrote:
>> I agree that the DDL behaviour is wrong and should be fixed. Thank you
>> for championing that alternative view.
>>
>> Swapping based upon names only works and is very flexible, muc
On Thu, Jan 20, 2011 at 11:17 AM, Simon Riggs wrote:
> On Wed, 2011-01-19 at 22:16 -0500, Robert Haas wrote:
>
>> That's another way of saying "the patch is not anywhere close to being done".
>
> My patch is materially incomplete. Certainly we may see that as grounds
> for rejection, which I would
On 21/01/11 15:24, Robert Haas wrote:
On Thu, Jan 20, 2011 at 9:17 PM, Kevin Grittner
wrote:
Right -- God only knows the number of things were filtered out to
leave me with filtered water. What's "filtered" in this case is what
was passed through, not what was removed.
Hmm, I guess I see you
On Thu, Jan 20, 2011 at 8:54 PM, Kevin Grittner
wrote:
> As I mentioned in another email, we might want to throttle this. My
> thinking was that we could start a timer on capturing a snapshot, and
> continue to gather new ones as they become available. When you hit
> the timer limit (maybe 100ms
"Kevin Grittner" writes:
>> Robert Haas wrote:
>> Oh, you mean water that had some things you didn't want taken out
>> of it?
> Right -- God only knows the number of things were filtered out to
> leave me with filtered water. What's "filtered" in this case is what
> was passed through, not wha
On Thu, Jan 20, 2011 at 9:17 PM, Kevin Grittner
wrote:
> Right -- God only knows the number of things were filtered out to
> leave me with filtered water. What's "filtered" in this case is what
> was passed through, not what was removed.
Hmm, I guess I see your point now. Well, I'm not wedded t
* Daniel Farina (drfar...@acm.org) wrote:
> On Thu, Jan 20, 2011 at 6:12 PM, Robert Haas wrote:
> > I'm not likely to write a patch for it, but if someone else writes one
> > I would be willing to (a) support it and (b) subject to consensus,
> > commit it.
>
> Wouldn't this require a client appli
> Robert Haas wrote:
> On Thu, Jan 20, 2011 at 3:57 PM, Kevin Grittner
> wrote:
>> Robert Haas wrote:
>>
>>> I think filtered is pretty clear and like it...
>>
>> I find it ambiguous. [Takes sip of filtered water.]
>
> Oh, you mean water that had some things you didn't want taken out
> of it?
On Jan21, 2011, at 03:14 , Daniel Farina wrote:
> On Thu, Jan 20, 2011 at 6:12 PM, Robert Haas wrote:
>> On Thu, Jan 20, 2011 at 9:07 PM, Stephen Frost wrote:
>>> * Robert Haas (robertmh...@gmail.com) wrote:
It strikes me that it would be useful to have a GUC that sets the
owner of any
On Thu, Jan 20, 2011 at 4:40 PM, Simone Aiken
wrote:
> After playing with this in benchmarks and researching the weird results I
> got I'm going to advise dropping the todo for now unless something happens
> to change how postgres handles clustering.
I agree, let's remove it.
That having been sa
On Jan21, 2011, at 03:03 , Robert Haas wrote:
> It strikes me that it would be useful to have a GUC that sets the
> owner of any new objects you create (much as you can control their
> default schemas using search_path). Obviously, you'd need to restrict
> it so that it wouldn't allow you to creat
On Thu, Jan 20, 2011 at 6:12 PM, Robert Haas wrote:
> On Thu, Jan 20, 2011 at 9:07 PM, Stephen Frost wrote:
>> * Robert Haas (robertmh...@gmail.com) wrote:
>>> It strikes me that it would be useful to have a GUC that sets the
>>> owner of any new objects you create (much as you can control their
On Thu, Jan 20, 2011 at 9:07 PM, Stephen Frost wrote:
> * Robert Haas (robertmh...@gmail.com) wrote:
>> It strikes me that it would be useful to have a GUC that sets the
>> owner of any new objects you create (much as you can control their
>> default schemas using search_path).
>
> There was a gre
On Thu, Jan 20, 2011 at 4:32 PM, Tom Lane wrote:
> Robert Haas writes:
>> On Thu, Jan 20, 2011 at 3:47 PM, Tom Lane wrote:
>>> BTW, is it just me, or is the terminology "number filtered" pretty
>>> confusing/ambiguous in itself? It doesn't seem at all clear to me
>>> whether that's the number o
* Robert Haas (robertmh...@gmail.com) wrote:
> It strikes me that it would be useful to have a GUC that sets the
> owner of any new objects you create (much as you can control their
> default schemas using search_path).
There was a great deal of discussion along these lines over the summer
of '09
On Thu, Jan 20, 2011 at 3:57 PM, Kevin Grittner
wrote:
> Robert Haas wrote:
>
>> I think filtered is pretty clear and like it...
>
> I find it ambiguous. [Takes sip of filtered water.]
Oh, you mean water that had some things you didn't want taken out of it?
--
Robert Haas
EnterpriseDB: http:/
On Thu, Jan 20, 2011 at 8:32 PM, Josh Berkus wrote:
>
>> * Eventual Retirement of old credentials without having to issue ALTER
>> statements (or really statements of any kind...) against application
>> schema objects.
>
> OK, that's a different goal. You want to be able to expire passwords
> wit
* Daniel Farina (drfar...@acm.org) wrote:
> I have thought about that, although LDAP is the only one that came to
> mind (I don't know a whole lot of systems in detail, only by name...so
> suggestions welcome for low-administrative-overhead variants).
My preference is Kerberos and I find that it w
On Thu, Jan 20, 2011 at 5:32 PM, Josh Berkus wrote:
>
>> * Eventual Retirement of old credentials without having to issue ALTER
>> statements (or really statements of any kind...) against application
>> schema objects.
>
> OK, that's a different goal. You want to be able to expire passwords
> wit
Tom Lane wrote:
> Simon Riggs writes:
>> On Wed, 2011-01-19 at 19:05 -0600, Kevin Grittner wrote:
>>> The idea is that whenever we see a valid snapshot which would
>>> yield a truly serializable view of the data for a READ ONLY
>>> transaction, we add a WAL record with that snapshot information.
> * Eventual Retirement of old credentials without having to issue ALTER
> statements (or really statements of any kind...) against application
> schema objects.
OK, that's a different goal. You want to be able to expire passwords
with an overlap period. That's quite different from wanting an
i
On Jan21, 2011, at 01:28 , Simon Riggs wrote:
> What I'm still not clear on is why that HS is different. Whatever rules
> apply on the master must also apply on the standby, immutably. Why is it
> we need to pass explicit snapshot information from master to standby? We
> don't do that, except at st
> What I'm still not clear on is why that HS is different. Whatever rules
> apply on the master must also apply on the standby, immutably. Why is it
> we need to pass explicit snapshot information from master to standby? We
> don't do that, except at startup for normal HS. Why do we need that?
>
>
Itagaki Takahiro wrote:
> On Fri, Jan 21, 2011 at 09:11, Bruce Momjian wrote:
> > What happened to our work to add a JSON data type for PG 9.1?
>
> Nothing will happen in 9.1.
> I assume we are in "competition" status:
>
> http://archives.postgresql.org/pgsql-hackers/2010-11/msg00481.php
>
> Al
On Fri, Jan 21, 2011 at 09:11, Bruce Momjian wrote:
> What happened to our work to add a JSON data type for PG 9.1?
Nothing will happen in 9.1.
I assume we are in "competition" status:
http://archives.postgresql.org/pgsql-hackers/2010-11/msg00481.php
Also, if PGXN will work well, we might not h
On Thu, Jan 20, 2011 at 4:35 PM, Josh Berkus wrote:
>
>> How does this work with newly created objects? Is there a way to have
>> them default objects to a different owner, the parent of the two
>> roles?
>
> No, but you could easily assign default permissions.
>
>> In the case of password rotatio
> How does this work with newly created objects? Is there a way to have
> them default objects to a different owner, the parent of the two
> roles?
No, but you could easily assign default permissions.
> In the case of password rotation, the goal would be to
> drop the old password after all clie
Tom Lane wrote:
> I'm pretty concerned about the performance implications, too. In
> particular that sounds like you could get an unbounded amount of
> WAL emitted from a *purely read only* transaction flow.
No. Read only transactions wouldn't create any flow at all. And I
suggested that we
On Fri, 2011-01-21 at 00:26 +0100, Florian Pflug wrote:
> On Jan21, 2011, at 00:11 , Simon Riggs wrote:
> > It's not clear to me what the reason is that this doesn't just work on
> > HS already. If you started there it might help.
>
>
> The problem is that snapshots taken on the master sometimes
On Thu, Jan 20, 2011 at 3:34 PM, Tom Lane wrote:
> Daniel Farina writes:
>> I wanted to test the waters on how receptive people might be to an
>> extension that would allow Postgres to support two passwords for a
>> given role.
>
> Not very. Why don't you just put two roles in the same group?
H
What happened to our work to add a JSON data type for PG 9.1?
--
Bruce Momjian http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
Simon Riggs writes:
> On Wed, 2011-01-19 at 19:05 -0600, Kevin Grittner wrote:
>> The idea is that whenever we see a valid snapshot which would yield
>> a truly serializable view of the data for a READ ONLY transaction,
>> we add a WAL record with that snapshot information.
> You haven't explain
Daniel Farina writes:
> I wanted to test the waters on how receptive people might be to an
> extension that would allow Postgres to support two passwords for a
> given role.
Not very. Why don't you just put two roles in the same group?
regards, tom lane
--
Sent via pgs
On 1/20/11 2:26 PM, Tom Lane wrote:
> Josh Berkus writes:
>>> I would take that to mean that it's waiting on the client.
>
>> You mean that the client timed out and isn't accepting data from the
>> query anymore?
>
> No, if the backend is in RECV state, it's waiting for the client to
> *send* it
I wrote:
> Why not? We already generate appropriate snapshots for this in
> SSI, so is the problem in getting the appropriate information into
> the WAL stream or in having a request for a snapshot within a
> serializable transaction while running in hot standby the problem?
I dropped few word
On Jan21, 2011, at 00:11 , Simon Riggs wrote:
> It's not clear to me what the reason is that this doesn't just work on
> HS already. If you started there it might help.
The problem is that snapshots taken on the master sometimes represent a
state of the database which cannot occur under any (vali
On 1/20/2011 12:47 PM, Tom Lane wrote:
So the line I'm thinking we should pursue is to visually associate the
new counter with the filter condition, either like
Filter Cond: (x> 42) (nfiltered = 123)
or
Filter Cond: (x> 42)
Rows Filtered: 123
I'd prefer the latter.
Simon Riggs wrote:
> On Wed, 2011-01-19 at 19:05 -0600, Kevin Grittner wrote:
>
>> The idea is that whenever we see a valid snapshot which would
>> yield a truly serializable view of the data for a READ ONLY
>> transaction, we add a WAL record with that snapshot information.
>
> You haven't exp
Robert Haas writes:
> On Thu, Jan 20, 2011 at 4:24 PM, Tom Lane wrote:
>> If you're willing to substitute an incompatible table, it's not clear
>> why you don't just do
>>
>>begin;
>>drop table t;
>>alter table t_new rename to t;
>>
On Thu, Jan 20, 2011 at 09:36:11PM +, Simon Riggs wrote:
> I agree that the DDL behaviour is wrong and should be fixed. Thank you
> for championing that alternative view.
>
> Swapping based upon names only works and is very flexible, much more so
> than EXCHANGE could be.
>
> A separate utili
Bosco Rama writes:
>>> If 'standard_conforming_strings = on' is set in our DB (which is required
>>> for
>>> our app) then the piped restore method (e.g. pg_restore -O backup.dat |
>>> psql)
>>> results in the large objects being corrupted.
> All servers and client tools involved are PG 8.4.6 o
On Wed, 2011-01-19 at 19:05 -0600, Kevin Grittner wrote:
> The idea is that whenever we see a valid snapshot which would yield
> a truly serializable view of the data for a READ ONLY transaction,
> we add a WAL record with that snapshot information.
You haven't explained why this approach is the
On Thu, Jan 20, 2011 at 2:45 PM, Andrew Dunstan wrote:
> Have you thought of trying to use an external auth source like LDAP for such
> a scheme?
I have thought about that, although LDAP is the only one that came to
mind (I don't know a whole lot of systems in detail, only by name...so
suggestion
On Thu, Jan 20, 2011 at 4:24 PM, Tom Lane wrote:
> Noah Misch writes:
>> Heikki's suggestion seemed straightforward, so much so that I couldn't figure
>> why nobody had done it. That would usually mean I'm missing something.
>
> If you're willing to substitute an incompatible table, it's not cle
On 01/20/2011 05:28 PM, Daniel Farina wrote:
Hello list,
I wanted to test the waters on how receptive people might be to an
extension that would allow Postgres to support two passwords for a
given role. I have recently encountered a case where this would be
highly useful when performing rollin
Hello list,
I wanted to test the waters on how receptive people might be to an
extension that would allow Postgres to support two passwords for a
given role. I have recently encountered a case where this would be
highly useful when performing rolling password upgrades across many
client applicatio
Josh Berkus writes:
>> I would take that to mean that it's waiting on the client.
> You mean that the client timed out and isn't accepting data from the
> query anymore?
No, if the backend is in RECV state, it's waiting for the client to
*send* it something.
(Although if this is an SSL connecti
Kevin,
> So, based on a more complete description of the issues, any more
> opinions on whether to generate the error, as suggested by Heikki?
If it's a choice between generating an error and letting users see
inconsistent data, I'll take the former.
> Does anyone think this justifies the compa
Dne 20.1.2011 11:05, Csaba Nagy napsal(a):
> Hi Tomas,
>
> On Wed, 2011-01-19 at 23:13 +0100, Tomas Vondra wrote:
>> No, the multi-column statistics do not require constant updating. There
>> are cases where a sampling is perfectly fine, although you may need a
>> bit larger sample. Generally if y
Heikki Linnakangas wrote:
> On 20.01.2011 03:05, Kevin Grittner wrote:
>> If we don't do something like this, do we just provide REPEATABLE
>> READ on the standby as the strictest level of transaction
>> isolation? If so, do we generate an error on a request for
>> SERIALIZABLE, warn and provide
Dne 20.1.2011 09:10, Heikki Linnakangas napsal(a):
> It seems that the suggested multi-column selectivity estimator would be
> more sensitive to ndistinct of the individual columns. Is that correct?
> How is it biased? If we routinely under-estimate ndistinct of individual
> columns, for example, d
> I would take that to mean that it's waiting on the client.
You mean that the client timed out and isn't accepting data from the
query anymore? Shouldn't Postgres time out on that after a while? In
one case, the orphaned statement was 16 hours old before we killed it.
If it's relevant, the cl
Dne 20.1.2011 03:36, Robert Haas napsal(a):
> On Wed, Jan 19, 2011 at 5:13 PM, Tomas Vondra wrote:
Regarding the crash scenario - if the commit fails, just throw away the
local estimator copy, it's not needed. I'm not sure how to take care of
the case when commit succeeds and the wr
After playing with this in benchmarks and researching the weird results I
got I'm going to advise dropping the todo for now unless something happens
to change how postgres handles clustering. You guys probably already
grokked this so I am just recording it for the list archives.
The primary
Josh Berkus writes:
> One of our clients is seeing an unusual issue with statements which are
> waiting going into sleep forever and never completing. This first
> e-mail is for a "has anyone else seen this problem before?" while we try
> to collect additional information for a diagnosis.
> An
On Thu, 2011-01-20 at 13:14 -0500, Noah Misch wrote:
> When DDL has taken AccessExclusiveLock and a query waits for it, it's the
> Right
> Thing for that query to wake up and proceed based on the complete, final state
> of that committed DDL. Aside from the waiting itself, the query should behav
> I have seen it -- on 8.1 too. On our case it was caused by an insert
> that was doing lots of toast insertions, so it needed to grab the
> extension lock frequently for the toast table; and this was slowed
> down by a largish shared_buffers setting, somehow (8.1 doesn't have lock
> partitioning
Robert Haas writes:
> On Thu, Jan 20, 2011 at 3:47 PM, Tom Lane wrote:
>> BTW, is it just me, or is the terminology "number filtered" pretty
>> confusing/ambiguous in itself? It doesn't seem at all clear to me
>> whether that's the number of rows passed by the filter condition or
>> the number o
Dne 20.1.2011 03:06, Nathan Boley napsal(a):
>> And actually it does not depend on ndistinct for the columns only, it
>> depends on ndistinct estimates for the combination of columns. So
>> improving the ndistinct estimates for columns is just a necessary first
>> step (and only if it works reasona
Noah Misch writes:
> Heikki's suggestion seemed straightforward, so much so that I couldn't figure
> why nobody had done it. That would usually mean I'm missing something.
If you're willing to substitute an incompatible table, it's not clear
why you don't just do
begin;
Excerpts from Josh Berkus's message of jue ene 20 18:05:15 -0300 2011:
>
> One of our clients is seeing an unusual issue with statements which are
> waiting going into sleep forever and never completing. This first
> e-mail is for a "has anyone else seen this problem before?" while we try
> to
Robert Haas writes:
> On Thu, Jan 20, 2011 at 2:10 PM, Dimitri Fontaine
> wrote:
>> Robert Haas writes:
>>> I think that the basic problem with wal_level is that to increase it
>>> you need to somehow ensure that all the backends have the new setting,
>>> and then checkpoint.
>>
>> Well, you ju
Hackers,
One of our clients is seeing an unusual issue with statements which are
waiting going into sleep forever and never completing. This first
e-mail is for a "has anyone else seen this problem before?" while we try
to collect additional information for a diagnosis.
This problem only happen
Robert Haas wrote:
> I think filtered is pretty clear and like it...
I find it ambiguous. [Takes sip of filtered water.] How about
excluded?
-Kevin
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpr
On Thu, Jan 20, 2011 at 3:47 PM, Tom Lane wrote:
> The main functional problem I see with this format is that it assumes
> there is one and only one filter step associated with every plan node.
> That is just plain wrong. Many don't have any, and there are important
> cases where there are two.
hubert depesz lubaczewski writes:
> On Thu, Jan 20, 2011 at 02:48:59PM -0500, Stephen Frost wrote:
>> He also mentioned that he didn't feel it was terribly complicated or
>> that it'd be difficult to update for this. Looking over the code, it's
>> got a simple regex for matching that line which w
On ons, 2011-01-19 at 10:06 +0900, Hitoshi Harada wrote:
> - This is not in the patch, but around line 184 "vis versa" in comment
> seems like typo.
Fixed.
> - A line break should be added before PLy_add_exception() after "static void"
I'll add that when I get to the patch.
> - This is also not
On lör, 2010-12-18 at 18:56 +0100, Jan Urbański wrote:
> there seems to be a problem in the way we add exceptions to the plpy
> module in PL/Python compiled with Python 3k.
>
> Try this: DO $$ plpy.SPIError $$ language plpython3u;
>
> I'm not a Python 3 expert, but I nicked some code from the Int
On 20.01.2011 22:15, Dimitri Fontaine wrote:
Robert Haas writes:
Sending them a signal seems like a promising approach, but the trick
is guaranteeing that they've actually acted on it before you start the
checkpoint.
How much using a latch here would help? Or be overkill?
A latch doesn't g
Robert Haas writes:
> Sending them a signal seems like a promising approach, but the trick
> is guaranteeing that they've actually acted on it before you start the
> checkpoint.
How much using a latch here would help? Or be overkill?
--
Dimitri Fontaine
http://2ndQuadrant.fr PostgreSQL : E
On Thu, Jan 20, 2011 at 02:48:59PM -0500, Stephen Frost wrote:
> * Robert Haas (robertmh...@gmail.com) wrote:
> > How much has that code been updated from one release to the next?
>
> Just an FYI, I talked to depesz on IRC (please chime in if you disagree
> with any of this) and he indicated that
On Thu, Jan 20, 2011 at 2:10 PM, Dimitri Fontaine
wrote:
> Robert Haas writes:
>> I think that the basic problem with wal_level is that to increase it
>> you need to somehow ensure that all the backends have the new setting,
>> and then checkpoint. Right now, the backends get the value through
>
On Thu, Jan 20, 2011 at 1:47 PM, Stephen Frost wrote:
> * Tom Lane (t...@sss.pgh.pa.us) wrote:
>> Robert Haas writes:
>> > I agree; we make bigger changes than this all the time.
>>
>> No, we don't.
>
> Alright, do we want to go down the road of adding new things to the
> XML/JSON/YAML/Whatever-e
* Robert Haas (robertmh...@gmail.com) wrote:
> How much has that code been updated from one release to the next?
Just an FYI, I talked to depesz on IRC (please chime in if you disagree
with any of this) and he indicated that he's had to update the code
from time to time, mostly because the parser
On Thu, Jan 20, 2011 at 09:26:29AM -0500, Robert Haas wrote:
> My main beef with the Boolean flags is that this kind of thing is not too
> clear:
>
>reindex_relation(myrel, false, false, true, true, false, true,
> false, false, true);
>
> Unless you have an excellent memory, you can't tell w
Robert Haas writes:
> I think that the basic problem with wal_level is that to increase it
> you need to somehow ensure that all the backends have the new setting,
> and then checkpoint. Right now, the backends get the value through
> the GUC machinery, and so there's no particular bound on how l
On 1/20/11 6:15 AM, Robert Haas wrote:
> On Thu, Jan 20, 2011 at 9:13 AM, Bruce Momjian wrote:
>> OK, I am ready to move test_fsync to /contrib. Is pg_test_fsync the
>> best name? pg_check_fsync? pg_fsync_performance? pg_verify_fsync?
>
> I don't see too much reason to rename it more than nec
* Tom Lane (t...@sss.pgh.pa.us) wrote:
> Robert Haas writes:
> > I agree; we make bigger changes than this all the time.
>
> No, we don't.
Alright, do we want to go down the road of adding new things to the
XML/JSON/YAML/Whatever-else format that isn't displayed in the TEXT
version, to avoid thi
On Thu, Jan 20, 2011 at 10:07:23AM +, Simon Riggs wrote:
> On Wed, 2011-01-19 at 17:46 -0500, Noah Misch wrote:
>
> > First, I'd like to note that the thread for this patch had *four* "me-too"
> > responses to the use case. That's extremely unusual; the subject is
> > definitely
> > compelli
On Jan 20, 2011 6:43 PM, "Tom Lane" wrote:
>
> Robert Haas writes:
> > On Thu, Jan 20, 2011 at 11:55 AM, Stephen Frost
wrote:
> >> While I agree completely about the general "if you're going to break,
> >> break it big" approach, but I don't particularly care for holding
output
> >> strings from
On Thu, Jan 20, 2011 at 12:57 PM, Magnus Hagander wrote:
>
> On Jan 20, 2011 6:43 PM, "Tom Lane" wrote:
>>
>> Robert Haas writes:
>> > On Thu, Jan 20, 2011 at 11:55 AM, Stephen Frost
>> > wrote:
>> >> While I agree completely about the general "if you're going to break,
>> >> break it big" appr
On Thu, Jan 20, 2011 at 11:59 AM, Dimitri Fontaine
wrote:
> Robert Haas writes:
>> Also, it won't actually work unless the server has replication
>> configured (wal_level!=minimal, max_wal_senders>0, and possibly some
>> setting for wal_keep_segments), which has been the main point of the
>> nami
Robert Haas writes:
> On Tue, Dec 21, 2010 at 10:49 PM, Robert Haas wrote:
>> On Tue, Dec 21, 2010 at 11:12 AM, Tom Lane wrote:
> Robert Haas writes:
> On Wed, Dec 15, 2010 at 11:29 AM, Tom Lane wrote:
- Writeable CTEs - I think we need Tom to pick this one up.
- Fix snapshot taking
1 - 100 of 148 matches
Mail list logo