Active connections are terminated because of small wal_sender_timeout

Hello Everyone!
Do you have any thoughts regarding this issue?

Best regards,
Andrei Yahorau

From: Andrei Yahorau/IBA
To: pgsql-general@postgresql.org,
Date: 21/06/2019 11:14
Subject: Active connections are terminated because of small
wal_sender_timeout

Hello PostgreSQL Community!

Not long ago I opened an issue regarding the problem about:
"terminating walsender process due to replication timeout" problem:
/messages/by-id/OF85C33E30.171C1C23-ON432583F9.003F5B16-432583F9.003FBAD7@iba.by

Thanks Everyone for your comprehensive explanations.
Basically I faced this problem because of small wal_sender_timeout value
(1 second) and quite large volume of data to be replicated.
There were some assumptions about some problems with my network. But
during my investigation I assured that it was not network problem.

I took a look in postgresql source code. As far as I understood walsender
can send some data to walreceiver regarding some changes and so-called
keepalive messages.
Exactly these keepalive messages walsender sends periodically once per
wal_sender_timeout seconds (once per 1 second in my case) and expects to
get responses from the walreceiver that everything goes ok.

I switched on trace and repeated my test. I found out that walreceiver
starts processing of the "keepalive" message in 3 seconds after its
sending.
As far as I understand it happened because walreceiver was busy by
accepting all the changes from the master and writing it to its standby
WAL logs.
So the standby postgres log was overloaded overfilled by the entries
regarding writing its own WALs:
18642 2019-05-28 05:08:31 EDT 00000 DEBUG: record known xact 559
latestObservedXid 559
18642 2019-05-28 05:08:31 EDT 00000 CONTEXT: WAL redo at 0/112CE9C0 for
Heap/LOCK: off 53: xid 559: flags 0 LOCK_ONLY EXCL_LOCK
18642 2019-05-28 05:08:31 EDT 00000 DEBUG: record known xact 559
latestObservedXid 559
18642 2019-05-28 05:08:31 EDT 00000 CONTEXT: WAL redo at 0/112CE9F8 for
Heap/UPDATE: off 53 xmax 559 ; new off 129 xmax 0
18642 2019-05-28 05:08:31 EDT 00000 DEBUG: record known xact 559
latestObservedXid 559
18642 2019-05-28 05:08:31 EDT 00000 CONTEXT: WAL redo at 0/112CEA48 for
Heap/LOCK: off 54: xid 559: flags 0 LOCK_ONLY EXCL_LOCK
18642 2019-05-28 05:08:31 EDT 00000 DEBUG: record known xact 559
latestObservedXid 559
18642 2019-05-28 05:08:31 EDT 00000 CONTEXT: WAL redo at 0/112CEA80 for
Heap/UPDATE: off 54 xmax 559 ; new off 130 xmax 0
18642 2019-05-28 05:08:31 EDT 00000 DEBUG: record known xact 559
latestObservedXid 559
So because of writing large volume of data it was not able to handle the
next messages quickly. It seems not to be related to network bandwidth or
CPU saturation.

Thereby, I see some kind of a contradiction with the official description
of wal_sender_timeout parameter:
Terminate replication connections that are inactive longer than the
specified number of milliseconds.
This is useful for the sending server to detect a standby crash or network
outage.

During my test the connection between master and standby was active and
there was no any network outage. So according to the description there was
no need to terminate replication connection.

So, I have some questions:
Is there any way (e. g. via configuration of other) to make the response
time to keepalive messages independent of the amount of data that the
walreceiver has to process?
If there is no such a way is it possible to update wal_sender_timeout
documentation so it reflects reality?

Best regards,
Andrei Yahorau

AYahorau@ibagroup.eu writes:

Do you have any thoughts regarding this issue?

I do not think anybody thinks this is a bug. Setting wal_sender_timeout
too small is a configuration mistake.

regards, tom lane

On Fri, Jul 05, 2019 at 10:03:16AM -0400, Tom Lane wrote:

AYahorau@ibagroup.eu writes:

Do you have any thoughts regarding this issue?

I do not think anybody thinks this is a bug. Setting wal_sender_timeout
too small is a configuration mistake.

Yeah. I don't see any bug here. Please note that it can be also a
problem to set up a too high value in some configuration setups. The
lack of flexibility in this area is why wal_sender_timeout has been
switch to be user-settable in v12. In short you can configure it in
the connection string to enforce a custom value per standby.
--
Michael

Hello Everyone!

I do not think anybody thinks this is a bug. Setting

wal_sender_timeout

too small is a configuration mistake.

I don't understand why it is a mistake. 1second is acceptable value for
wal_sender_timeout.
Moreover the behaviour contradicts with the official description for
wal_sender_timeout:
Terminate replication connections that are inactive longer than the
specified number of milliseconds.

First of all the connection between the master and standby was good in
my example. The problem was in keepalive message processing (because of
busy standby server).
So, here 2 variants are possible:
1). Inappropriate keepalive message processing which contradicts with
wal_sender_timeout description.
2) Incorrect description of wal_sender_timeout parameter in the
documentation.

Yeah. I don't see any bug here. Please note that it can be also a

problem to set up a too high value in some configuration setups. The lack
of flexibility in this area is why wal_sender_timeout has been switch to
be

user-settable in v12. In short you can configure it in the connection

string to enforce a custom value per standby.
Thanks for this announcement, This enhancement looks very useful.

Best regards,
Andrei Yahorau

Hello Everyone!

Sorry for being persistent.

I do not think anybody thinks this is a bug. Setting wal_sender_timeout
too small is a configuration mistake.

Why is it a configuration mistake? This value is allowed to be set. There
is no any restriction about it.
I would like to ask a question regarding wal_sender_timeout description
and its real behaviour.
Description:
Terminate replication connections that are inactive longer than the
specified number of milliseconds.
Real behaviour:
A standby server can be busy and does not send "keepalive" packets to the
master.
(By the way wal_sender_timeot cannot be not so small as in my tests. This
situation can happen when wal_sender_timeout=10seconds and so on and and
so forth).

Does it work properly according to the description? Don't you see the
contradiction between the description and real behaviour? As I mentioned
before the connection between master and standby is good. There is no any
problem.
So where is a bug: in the description or in the implementation?

Yeah. I don't see any bug here. Please note that it can be also a
problem to set up a too high value in some configuration setups. The
lack of flexibility in this area is why wal_sender_timeout has been
switch to be user-settable in v12. In short you can configure it in
the connection string to enforce a custom value per standby.

Will a small value of wal_sender_timeout in PostgreSQL v12 lead to the
same failure "terminating walsender process due to replication timeout" as
we observe in v11 ?

Best regards,
Andrei Yahorau

AYahorau@ibagroup.eu writes:

I do not think anybody thinks this is a bug. Setting wal_sender_timeout
too small is a configuration mistake.

Why is it a configuration mistake? This value is allowed to be set. There
is no any restriction about it.

The fact that a parameter can be set does not mean that we guarantee that
every possible value will work for everybody. As an example, if you
configure work_mem to a few TB and then start getting OOM failures because
your machine can't actually support that, it's not a bug that we let you
set the value that high. The upper limit of what's usable is too variable
and too hard to determine, so we don't even try; it's on you to choose a
suitable setting for your situation. Similarly, the useful range of
wal_sender_timeout is hard to predict and is likely to be quite different
for different installations. We don't want to artificially constrain what
people can use, so the range of allowed settings *ought* to include some
values that are not practically useful in specific situations. Those
values might be just the right ones for someone else.

regards, tom lane