On Mon, 8 Sep 2008, Alvaro Herrera [EMAIL PROTECTED] writes:
Modified as you suggested. BTW, will there be a similar i18n scenario
for dropped column you mentioned below?
Yes, you need _() around those too.
For this purpose, I introduced a dropped_column_type variable in
Fujii Masao [EMAIL PROTECTED] wrote:
3) Patch of introducing new background process which I've called
WALSender. It takes charge of sending WAL to the slave.
Now, I assume that WALSender also listens the connection from
the slave, i.e. only one sender process manages
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't think there's a easy way to fix this. Maybe
we should not allow WITH clause without RECURISVE?
My
Hi,
I am currently working in EnterpriseDB (www.enterprisedb.com). During the
integration of pldebugger ( http://pgfoundry.org/projects/edb-debugger ) with
postgres on windows I faced a problem that plugins are not being copied to the
lib/plugins directory. Plugins should be copied in
On Mon, 2008-09-08 at 17:40 -0400, Bruce Momjian wrote:
Fujii Masao wrote:
On Mon, Sep 8, 2008 at 8:44 PM, Markus Wanner [EMAIL PROTECTED] wrote:
Merge into WAL writer?
Uh.. that would mean you'd loose parallelism between WAL writing to disk
and
WAL shipping via network.
On Mon, 8 Sep 2008, Bruce Momjian wrote:
I don't think much changes at the flow chart level from release to
release so it would fine if it was just CVS HEAD. I also don't think
many people do back-branch development.
It's easy with Mediawiki to both 1) see the old version for those
Alvaro Herrera wrote:
Tom Lane wrote:
Alvaro Herrera [EMAIL PROTECTED] writes:
Tom Lane wrote:
Hmm, I didn't recheck after Greg's patch, but in mine, it doesn't,
because the location is saved as reset location and restored when the
variable is reset. It worked fine in all cases I tested.
On Sun, 7 Sep 2008, M2Y wrote:
Why does a replication solution need log shipping and why cant we just
ship the transaction statements to a standby node?
Here's one of the classic examples of why that doesn't work:
create table x (d decimal);
insert into x values (random());
If you execute
Simon Riggs wrote:
This gives us the Group Commit feature also, even if we are not using
replication. So we can drop the commit_delay stuff.
Huh? How does that give us group commit?
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com
--
Sent via pgsql-hackers mailing list
On Tue, 9 Sep 2008, Magnus Hagander wrote:
(I dropped the default stuff for now, as it doesn't seem that a
consensus has been reached on that topic.)
This is one of the reasons I suggested keeping that one as a separate
patch in the first place. The other main reason being that once it gets
On Mon, Sep 08, 2008 at 06:24:14PM +1000, Brendan Jurd wrote:
On Sun, Sep 7, 2008 at 5:58 AM, Alex Hunsaker [EMAIL PROTECTED] wrote:
Im just following this:
http://wiki.postgresql.org/wiki/Reviewing_a_Patch so lets get started.
Hi Alex. Thanks for taking the time to review my patch.
I
On Tue, 2008-09-09 at 12:24 +0300, Heikki Linnakangas wrote:
Simon Riggs wrote:
This gives us the Group Commit feature also, even if we are not using
replication. So we can drop the commit_delay stuff.
Huh? How does that give us group commit?
Multiple backends waiting while we perform a
Simon Riggs wrote:
Multiple backends waiting while we perform a write. Commits then happen
as a group (to WAL at least), hence Group Commit.
The problem with our current commit protocol is this:
1. Backend A inserts commit record A
2. Backend A starts to flush commit record A
3. Backend B
Tom Lane [EMAIL PROTECTED] writes:
Gregory Stark [EMAIL PROTECTED] writes:
Yeah, I've been thinking about how to use the planner to do this.
I thought the answer to that was going to be more or less call
cost_sort() and cost_index() and compare the answers.
That was the way I was headed.
Heikki Linnakangas [EMAIL PROTECTED] wrote:
The tricky part is, how does A know if it should wait, and for how long?
commit_delay sure isn't ideal, but AFAICS the log shipping proposal
doesn't provide any solution to that.
They have no relation each other directly,
but they need similar
On Tue, Sep 9, 2008 at 7:29 PM, Martijn van Oosterhout
[EMAIL PROTECTED] wrote:
I actually had a look at this patch also, though not as thoroughly as
Alex. There was one part that I had some thoughts about in
from_char_parse_int_len:
Hi Martijn. Thanks for your comments.
The use of
On Tue, Sep 9, 2008 at 5:11 PM, Simon Riggs [EMAIL PROTECTED] wrote:
Yes. We should have a LogwrtRqst pointer and LogwrtResult pointer for
the send operation. The Write and Send operations can then continue
independently of one another. XLogInsert() cannot advance to a new page
while we are
Hi,
ITAGAKI Takahiro wrote:
Signals and locking, borrewed from Postgres-R, are now studied
for the purpose in the log shipping, but I'm not sure it can be
also used in the group commit.
Yeah. As Heikki points out, there is a completely orthogonal question
WRT group commit: how does
Hi,
Fujii Masao wrote:
Really? In the benchmark result of my prototype, the bottleneck is
still disk I/O.
The communication (between the master and the slave) latency is smaller than
WAL writing (fsyncing) one. Of course, I assume that we use not-poor network
like 1000BASE-T.
Sure. If you do
On Tue, 2008-09-09 at 20:12 +0900, Fujii Masao wrote:
What makes the sender process bottleneck?
In my experience, the Atlantic. But I guess the Pacific does it too. :-)
--
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support
--
Sent via pgsql-hackers
Fujii Masao wrote:
What makes the sender process bottleneck?
The keyword here is might. There's many possibilities, like:
- Slow network.
- Ridiculously fast disk. Like a RAM disk. If you have a synchronous
slave you can fail over to, putting WAL on a RAM disk isn't that crazy.
- slower WAL
On Tue, 2008-09-09 at 12:54 +0300, Heikki Linnakangas wrote:
Note that we already have the logic to flush all pending commit
records at once.
But only when you can grab WALInsertLock when flushing. If you look at
the way I suggested, it does not rely upon that lock being available.
So it is
On Tue, 2008-09-09 at 20:12 +0900, Fujii Masao wrote:
I'd like to introduce new parameter synchronous_replication which specifies
whether backends waits for the response from WAL sender process. By
combining synchronous_commit and synchronous_replication, users can
choose various options.
On Tue, 2008-09-09 at 13:42 +0200, Markus Wanner wrote:
How about multiple standby servers?
There are various ways for getting things to work with multiple servers.
I hope we can make this work with just a single standby before we try to
make it work on more. There are various options for
Gregory Stark [EMAIL PROTECTED] writes:
In particular I'm thinking of people clustering on a covering index (which
isn't as uncommon as it sounds, if you have a covering index you probably do
want to cluster it -- consider many-to-many join tables). We should be able to
do an index-only scan
On Tue, 2008-09-09 at 08:24 -0400, Tom Lane wrote:
Fujii Masao [EMAIL PROTECTED] writes:
On Tue, Sep 9, 2008 at 5:11 PM, Simon Riggs [EMAIL PROTECTED] wrote:
Yes. We should have a LogwrtRqst pointer and LogwrtResult pointer for
the send operation. The Write and Send operations can then
Fujii Masao [EMAIL PROTECTED] writes:
On Tue, Sep 9, 2008 at 5:11 PM, Simon Riggs [EMAIL PROTECTED] wrote:
Yes. We should have a LogwrtRqst pointer and LogwrtResult pointer for
the send operation. The Write and Send operations can then continue
independently of one another. XLogInsert()
On Tue, Sep 9, 2008 at 9:04 PM, Brendan Jurd [EMAIL PROTECTED] wrote:
On Tue, Sep 9, 2008 at 7:29 PM, Martijn van Oosterhout
[EMAIL PROTECTED] wrote:
The use of palloc/pfree in this routine seems excessive. Does len have
upper bound? If so a simple array will do it.
I suppose I could define
Simon Riggs [EMAIL PROTECTED] writes:
On Tue, 2008-09-09 at 08:24 -0400, Tom Lane wrote:
Agreed? That last restriction is a deal-breaker.
OK, I should have said *if wal_buffers are full* XLogInsert() cannot
advance to a new page while we are waiting to send or write. So I don't
think its a
Hi,
Le mardi 09 septembre 2008, Heikki Linnakangas a écrit :
The tricky part is, how does A know if it should wait, and for how long?
commit_delay sure isn't ideal, but AFAICS the log shipping proposal
doesn't provide any solution to that.
It might just be I'm not understanding what it's all
Greg Smith wrote:
On Tue, 9 Sep 2008, Magnus Hagander wrote:
(I dropped the default stuff for now, as it doesn't seem that a
consensus has been reached on that topic.)
This is one of the reasons I suggested keeping that one as a separate
patch in the first place. The other main reason being
Alvaro Herrera wrote:
Attached is a patch that implements the described signalling.
Are we keeping the idea that the reader sends a stat message when it
needs to read stats? What about the lossiness of the transport?
As the message is resent in the wait loop, the collector should receive
Hi,
Dimitri Fontaine wrote:
It might just be I'm not understanding what it's all about, but it seems to me
with WALSender process A will wait, whatever happens, either until the WAL is
sent to slave or written to disk on the slave.
..and it will still has to wait until WAL is written to disk
Le mardi 09 septembre 2008, Markus Wanner a écrit :
..and it will still has to wait until WAL is written to disk on the
local node, as we do now. These are two different things to wait for.
One is a network socket operation, the other is an fsync(). As these
don't work together too well
On Tue, 2008-09-09 at 15:32 +0200, Dimitri Fontaine wrote:
The process is now already waiting in all cases
If the WALWriter|Sender is available, it can begin the task immediately.
There is no need for it to wait if you want synchronous behaviour.
--
Simon Riggs www.2ndQuadrant.com
3. I think this is a must fix because of the point about volatile
functions --- changing it later will result in user-visible semantics
changes, so we have to get it right the first time.
I don't entirely agree with #3. It is user-visible, but only in the
sense that someone is depending on
Hi,
ITAGAKI Takahiro wrote:
Signals and locking, borrewed from Postgres-R, are now studied
for the purpose in the log shipping,
Cool. Let me know if you have any questions WRT this imessages stuff.
Regards
Markus Wanner
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
Hi,
Dimitri Fontaine wrote:
Exactly the point. The process is now already waiting in all cases, so maybe
we could just force waiting some WALSender signal before sending the fsync()
order, so we now have Group Commit.
A single process can only wait on either fsync() or on select(), but not
Le mardi 09 septembre 2008, Simon Riggs a écrit :
If the WALWriter|Sender is available, it can begin the task immediately.
There is no need for it to wait if you want synchronous behaviour.
Ok. Now I'm as lost as anyone with respect to how you get Group Commit :)
--
dim
signature.asc
On Tue, 2008-09-09 at 16:05 +0200, Dimitri Fontaine wrote:
Le mardi 09 septembre 2008, Simon Riggs a écrit :
If the WALWriter|Sender is available, it can begin the task immediately.
There is no need for it to wait if you want synchronous behaviour.
Ok. Now I'm as lost as anyone with
Tom Lane wrote:
Simon Riggs [EMAIL PROTECTED] writes:
On Tue, 2008-09-09 at 08:24 -0400, Tom Lane wrote:
Agreed? That last restriction is a deal-breaker.
OK, I should have said *if wal_buffers are full* XLogInsert() cannot
advance to a new page while we are waiting to send or write. So I
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't think there's a easy way to fix this. Maybe
we should not allow WITH clause without RECURISVE?
Hello
2008/9/9 Tatsuo Ishii [EMAIL PROTECTED]:
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't think there's a easy way to fix this. Maybe
we
Hello
2008/9/9 Tatsuo Ishii [EMAIL PROTECTED]:
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't think there's a easy way to fix this. Maybe
Hello
2008/9/9 Tatsuo Ishii [EMAIL PROTECTED]:
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't think there's a easy way to fix this. Maybe
On Tue, 2008-09-09 at 17:17 +0300, Heikki Linnakangas wrote:
Tom Lane wrote:
Simon Riggs [EMAIL PROTECTED] writes:
On Tue, 2008-09-09 at 08:24 -0400, Tom Lane wrote:
Agreed? That last restriction is a deal-breaker.
OK, I should have said *if wal_buffers are full* XLogInsert() cannot
* Aggregates allowed:
with recursive foo(i) as
(values(1)
union all
select max(i)+1 from foo where i 10)
select * from foo;
Aggregates should be blocked according to the standard.
Also, causes an infinite loop. This should be fixed for 8.4.
I will try
Volkan YAZICI wrote:
On Mon, 8 Sep 2008, Alvaro Herrera [EMAIL PROTECTED] writes:
Modified as you suggested. BTW, will there be a similar i18n scenario
for dropped column you mentioned below?
Yes, you need _() around those too.
For this purpose, I introduced a dropped_column_type
Simon Riggs wrote:
Don't understand. I am referring to the logic at the top of
AdvanceXLInsertBuffer(). We would need to wait for all people reading
the contents of wal_buffers.
Oh, I see.
If a slave falls behind, how does it catch up? I guess you're saying
that it can't fall behind,
On Tue, 2008-09-09 at 18:26 +0300, Heikki Linnakangas wrote:
Simon Riggs wrote:
Don't understand. I am referring to the logic at the top of
AdvanceXLInsertBuffer(). We would need to wait for all people reading
the contents of wal_buffers.
Oh, I see.
If a slave falls behind, how does
2008/9/9 Tatsuo Ishii [EMAIL PROTECTED]:
Hello
2008/9/9 Tatsuo Ishii [EMAIL PROTECTED]:
On Tue, 2008-09-09 at 13:45 +0900, Tatsuo Ishii wrote:
Thanks for the review.
The standard specifies that non-recursive WITH should be evaluated
once.
What shall we do? I don't
My interpretation of 7.13: General Rules: 2.b is that it should be
single evaluation, even if RECURSIVE is present.
The previous discussion was here:
http://archives.postgresql.org/pgsql-hackers/2008-07/msg01292.php
I am blind, I didn't find any reason, why materialisation isn't
Also, current implementation has only a type of plan which uses sort
operation. It should be optimized by re-position the windows and/or
using hashtable.
I would like to see some performance test results also. It would be good
to know whether they are fast/slow etc.. It will definitely help
2008/9/9 Robert Haas [EMAIL PROTECTED]:
My interpretation of 7.13: General Rules: 2.b is that it should be
single evaluation, even if RECURSIVE is present.
The previous discussion was here:
http://archives.postgresql.org/pgsql-hackers/2008-07/msg01292.php
I am blind, I didn't find
Tom Lane wrote:
Greg Stark [EMAIL PROTECTED] writes:
On Mon, Sep 8, 2008 at 2:11 PM, Tom Lane [EMAIL PROTECTED] wrote:
But of course case insensitivity isn't going to fix that example for you.
So we're right back at the question of where we should draw the line in
trying to accept variant
On Tue, 2008-09-09 at 18:51 +0200, Pavel Stehule wrote:
hmm. I solve similar problem in grouping sets :( etc
How did you solve it?
Regards,
Jeff Davis
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
2008/9/9 Jeff Davis [EMAIL PROTECTED]:
On Tue, 2008-09-09 at 18:51 +0200, Pavel Stehule wrote:
hmm. I solve similar problem in grouping sets :( etc
I have special executor node - feeder, it hold one tuple and others
nodes read from this node. It's usable for hash aggregates.
Pavel
plan is
Peter Eisentraut wrote:
There is interest among packagers to run the regression tests or other
tests after the build. The Red Hat RPMs have shipped a postgresql-test
package for years with a hacked-up makefile that will probably overwrite
random files that it shouldn't in /usr/lib. So
PostgreSQL development community:
Our research group has been using the PostgreSQL code base to test new
join algorithms. During testing, we noticed that the planner is not
pushing down projections to the outer relation in a hash join. Although
this makes sense for in-memory (1 batch) joins,
Don't understand. I am referring to the logic at the top of
AdvanceXLInsertBuffer(). We would need to wait for all people reading
the contents of wal_buffers.
Oh, I see.
If a slave falls behind, how does it catch up? I guess you're saying
that it can't fall behind, because the master
Martin Pihlak [EMAIL PROTECTED] writes:
Yes, creating a new message type was a bit short sighted -- attached is a
patch
that uses syscache invalidation messages instead. This also adds additional
tupleId field to SharedInvalCatcacheMsg. This is used to identify the
invalidated tuple in PROC
Hi,
AFAICS, PostgreSQL is not keeping info about when a table, database,
sequence, etc was created. We cannot get that info even from OS, since
CLUSTER or VACUUM FULL may change the metadata of corresponding
relfilenode.
Does anyone think that adding a timestamp column to pg_class would bring
an
This would have definitional problems of its own, however. If you interpret
K, M, and G strictly as unit-less multipliers, then
SET shared_buffers = 2 G
I don't think it would be a good idea to make them unit-less, for
exactly the reasons you mention.
We could possibly settle some of these
Devrim =?ISO-8859-1?Q?G=DCND=DCZ?= [EMAIL PROTECTED] writes:
AFAICS, PostgreSQL is not keeping info about when a table, database,
sequence, etc was created. We cannot get that info even from OS, since
CLUSTER or VACUUM FULL may change the metadata of corresponding
relfilenode.
Does anyone
On Tue, Sep 09, 2008 at 03:36:19PM -0400, Tom Lane wrote:
Devrim =?ISO-8859-1?Q?G=DCND=DCZ?= [EMAIL PROTECTED] writes:
AFAICS, PostgreSQL is not keeping info about when a table, database,
sequence, etc was created. We cannot get that info even from OS, since
CLUSTER or VACUUM FULL may
On Tue, 2008-09-09 at 15:36 -0400, Tom Lane wrote:
why would creation time (as opposed to any other time, eg last schema
modification, last data modification, yadda yadda) be especially
significant?
Hmm, those would be cool, too.
Seriously, I believe we can get last data modification from
There isn't sufficient support for such a feature.
It sounds like a useful feature to me.
In any case, why
would creation time (as opposed to any other time, eg last schema
modification, last data modification, yadda yadda) be especially
significant?
Those would be nice to have too, but
On Tue, 2008-09-09 at 12:40 -0700, daveg wrote:
I'd be very interested in seeing a last schema modification time for pg_class
objects. I don't care about it being preserved over dump and restore as my
use case is more to find out when a table was created with a view to finding
out if it is
Robert Haas [EMAIL PROTECTED] writes:
I am blind, I didn't find any reason, why materialisation isn't useable.
I believe it's because of these two (closely related) problems:
# The basic
# implementation clearly ought to be to dump the result of the subquery
# into a tuplestore and then
On Tue, Sep 09, 2008 at 11:03:56PM +0300, Hannu Krosing wrote:
On Tue, 2008-09-09 at 12:40 -0700, daveg wrote:
I'd be very interested in seeing a last schema modification time for
pg_class
objects. I don't care about it being preserved over dump and restore as my
use case is more to
daveg wrote:
On Tue, Sep 09, 2008 at 11:03:56PM +0300, Hannu Krosing wrote:
On Tue, 2008-09-09 at 12:40 -0700, daveg wrote:
I'd be very interested in seeing a last schema modification time for pg_class
objects. I don't care about it being preserved over dump and restore as my
use
Hi Andrew,
On Tue, 2008-09-09 at 16:22 -0400, Andrew Dunstan wrote:
I'd like to know why logging statements won't do the trick here.
It is not on by default, logs are rotated, and may be lost, etc.
Regards,
--
Devrim GÜNDÜZ, RHCE
devrim~gunduz.org, devrim~PostgreSQL.org,
On Thu, 2008-09-04 at 11:51 -0400, Andrew Sullivan wrote:
On Thu, Sep 04, 2008 at 07:01:18AM -0700, Steve Atkins wrote:
Settings in postgresql.conf are currently case-insensitive. Except
for the units.
And, of course, filenames when you are using a case-sensitive
filesystem. Because
On Tue, 9 Sep 2008, Robert Haas wrote:
A good start might be to always OUTPUT memory parameters using the
same base unit.
SHOW gives output that matches what you input. If you want to see things
with consistant units, look at pg_settings:
# select name,unit,setting,current_setting(name)
On Tue, Sep 09, 2008 at 10:20:00PM +0300, Devrim GUNDUZ wrote:
Hi,
AFAICS, PostgreSQL is not keeping info about when a table, database,
sequence, etc was created. We cannot get that info even from OS,
since CLUSTER or VACUUM FULL may change the metadata of
corresponding relfilenode.
When
Over in that TPC-H thread, I was bemoaning once again the never-finished
support for SQL-spec interval literals. I decided to go look at exactly
how unfinished it was, and it turns out that it's actually pretty close.
Hence the attached proposed patch ;-)
The main gating factor is that
Andrew Dunstan wrote:
...
Can someone please give a good, concrete use case for this stuff? Might
be nice to have doesn't cut it, I'm afraid. In particular, I'd like to
know why logging statements won't do the trick here.
Please pardon the kibbitzer intrusion ...
Informix has this
Greg Smith [EMAIL PROTECTED] writes:
What I would like to see (but don't
have nearly enough time to argue in support of considering the resistance
to change here) is that this syntax:
shared_buffers=1024
Would assume the user meant 1024 *bytes*, with the server silently
rounding that up
On Mon, Sep 8, 2008 at 2:24 AM, Brendan Jurd [EMAIL PROTECTED] wrote:
HEAD actually gets this one wrong; in defiance of the documentation it
returns 2000-09-07. So, it seems to me that the patch shifts the
behaviour in the right direction.
Barring actually teaching the code that some nodes
Alvaro Herrera [EMAIL PROTECTED] writes:
Peter Eisentraut wrote:
There is interest among packagers to run the regression tests or other
tests after the build. The Red Hat RPMs have shipped a postgresql-test
package for years with a hacked-up makefile that will probably overwrite
random
On Tue, Sep 9, 2008 at 6:46 AM, Brendan Jurd [EMAIL PROTECTED] wrote:
On Tue, Sep 9, 2008 at 9:04 PM, Brendan Jurd [EMAIL PROTECTED] wrote:
On Tue, Sep 9, 2008 at 7:29 PM, Martijn van Oosterhout
[EMAIL PROTECTED] wrote:
The use of palloc/pfree in this routine seems excessive. Does len have
Lawrence, Ramon [EMAIL PROTECTED] writes:
Our research group has been using the PostgreSQL code base to test new
join algorithms. During testing, we noticed that the planner is not
pushing down projections to the outer relation in a hash join. Although
this makes sense for in-memory (1
On Tue, 9 Sep 2008, Tom Lane wrote:
How do you come to that conclusion? Leaving off the unit entirely
certainly doesn't make the user's intent clearer.
Same way I do all my conclusions in this area--talking to people in the
field regularly who've never configured a postgresql.conf before.
A good start might be to always OUTPUT memory parameters using the
same base unit.
SHOW gives output that matches what you input.
Not for me it doesn't.
portal=# show temp_buffers;
temp_buffers
--
1024
(1 row)
portal=# set temp_buffers = '16MB';
SET
portal=# show temp_buffers;
When people aren't keeping track of their DDL, that is very strictly a
process problem on their end. When people are shooting themselves in
the foot, it's a great disservice to market Kevlar shoes to them.
I can't believe anyone is going to stop tracking their DDL because,
ooh goody, now we
Robert Haas wrote:
I try pretty hard not to shoot myself in the foot. But if someone
comes up to me and offers me some shoes that are have the same cost,
appearance, comfort-level, and durability as regular shoes but are
slightly more bullet resistant, should I refuse them on principle?
Why?
On Tue, Sep 9, 2008 at 8:38 PM, Heikki Linnakangas
[EMAIL PROTECTED] wrote:
There's one thing I haven't figured out in this discussion. Does the write
to the disk happen before or after the write to the slave? Can you guarantee
that if a transaction is committed in the master, it's also
On Tue, Sep 9, 2008 at 8:42 PM, Markus Wanner [EMAIL PROTECTED] wrote:
In the viewpoint of detection of a network failure, this feature is
necessary.
When the network goes down, WAL sender can be blocked until it detects
the network failure, i.e. WAL sender keeps waiting for the response which
88 matches
Mail list logo