Horton

Re: Minimal logical decoding on standbys

Amit Khandekar <amitdkhan.pg@gmail.com>

View thread
From: Amit Khandekar <amitdkhan.pg@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: tushar <tushar.ahuja@enterprisedb.com>, Petr Jelinek <petr.jelinek@2ndquadrant.com>, Robert Haas <robertmhaas@gmail.com>, Craig Ringer <craig@2ndquadrant.com>, Petr Jelinek <petr@2ndquadrant.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2019-04-16T06:57:46Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Reduce the log level in 035_standby_logical_decoding.pl.

  2. 035_standby_logical_decoding: Add missing waits for replication

  3. For cascading replication, wake physical and logical walsenders separately

  4. Handle logical slot conflicts on standby

  5. Support invalidating replication slots due to horizon and wal_level

  6. Prevent use of invalidated logical slot in CreateDecodingContext()

  7. Replace replication slot's invalidated_at LSN with an enum

  8. Pass down table relation into more index relation functions

  9. Assert only valid flag bits are passed to visibilitymap_set()

  10. Remove unused _bt_delitems_delete() argument.

  11. Add xl_btree_delete optimization.

On Sat, 13 Apr 2019 at 00:57, Andres Freund <andres@anarazel.de> wrote:
>
> Hi,
>
> On 2019-04-12 23:34:02 +0530, Amit Khandekar wrote:
> > I tried to see if I can quickly understand what's going on.
> >
> > Here, master wal_level is hot_standby, not logical, though slave
> > wal_level is logical.
>
> Oh, that's well diagnosed.  Cool.  Also nicely tested - this'd be ugly
> in production.

Tushar had made me aware of the fact that this reproduces only when
master wal_level is hot_standby.

>
> I assume the problem isn't present if you set the primary to wal_level =
> logical?

Right.

>
>
> > Not sure why this is happening. On slave, wal_level is logical, so
> > logical records should have tuple data. Not sure what does that have
> > to do with wal_level of master. Everything should be there on slave
> > after it replays the inserts; and also slave wal_level is logical.
>
> The standby doesn't write its own WAL, only primaries do. I thought we
> forbade running with wal_level=logical on a standby, when the primary is
> only set to replica.  But that's not what we do, see
> CheckRequiredParameterValues().
>
> I've not yet thought this through, but I think we'll have to somehow
> error out in this case.  I guess we could just check at the start of
> decoding what ControlFile->wal_level is set to,

By "start of decoding", I didn't get where exactly. Do you mean
CheckLogicalDecodingRequirements() ?

> and then raise an error
> in decode.c when we pass an XLOG_PARAMETER_CHANGE record that sets
> wal_level to something lower?

Didn't get where exactly we should error out. We don't do
XLOG_PARAMETER_CHANGE handling in decode.c , so obviously you meant
something else, which I didn't understand.

What I am thinking is :
In CheckLogicalDecodingRequirements(), besides checking wal_level,
also check ControlFile->wal_level when InHotStandby. I mean, when we
are InHotStandby, both wal_level and ControlFile->wal_level should be
>= WAL_LEVEL_LOGICAL. This will allow us to error out when using logical
slot when master has incompatible wal_level.

ControlFile is not accessible outside xlog.c so need to have an API to
extract this field.


>
> Could you try to implement that?
>
> Greetings,
>
> Andres Freund


--
Thanks,
-Amit Khandekar
EnterpriseDB Corporation
The Postgres Database Company