9.2 Cascading replication after slave promotion

Started by Mark Kirkwoodover 13 years ago7 messages
#1Mark Kirkwood
mark.kirkwood@catalyst.net.nz

Looking at the docs (section 25.2.6 paragraph 6) leads one to believe
that downstream standbys can continue to receive and process wal merely
by reconnecting after the cascading standby is promoted - but this does
not seem to be the case (indeed the same paragraph notes that timelines
now no longer match). It looks to me like the downstream guys all need
to be rebuilt - or am I missing something?

Regards

mark

#2Josh Berkus
josh@agliodbs.com
In reply to: Mark Kirkwood (#1)
Re: 9.2 Cascading replication after slave promotion

Mark,

Looking at the docs (section 25.2.6 paragraph 6) leads one to believe
that downstream standbys can continue to receive and process wal merely
by reconnecting after the cascading standby is promoted - but this does
not seem to be the case (indeed the same paragraph notes that timelines
now no longer match). It looks to me like the downstream guys all need
to be rebuilt - or am I missing something?

Per discussion on this list (search for "remastering"), 9.2 still won't
allow standbies to point to a new master unless you are doing file-based
replication.

To date, I seem to be the only one convinced that this is an actual
deficiency ...

--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com

#3Mark Kirkwood
mark.kirkwood@catalyst.net.nz
In reply to: Josh Berkus (#2)
Re: 9.2 Cascading replication after slave promotion

On 14/08/12 12:32, Josh Berkus wrote:

Mark,

Looking at the docs (section 25.2.6 paragraph 6) leads one to believe
that downstream standbys can continue to receive and process wal merely
by reconnecting after the cascading standby is promoted - but this does
not seem to be the case (indeed the same paragraph notes that timelines
now no longer match). It looks to me like the downstream guys all need
to be rebuilt - or am I missing something?

Per discussion on this list (search for "remastering"), 9.2 still won't
allow standbies to point to a new master unless you are doing file-based
replication.

To date, I seem to be the only one convinced that this is an actual
deficiency ...

Make that 2 of us...

#4Daniel Farina
daniel@heroku.com
In reply to: Josh Berkus (#2)
Re: 9.2 Cascading replication after slave promotion

On Mon, Aug 13, 2012 at 5:32 PM, Josh Berkus <josh@agliodbs.com> wrote:

To date, I seem to be the only one convinced that this is an actual
deficiency ...

It is an actual deficiency, and I am convinced.

--
fdr

#5Stephen Frost
sfrost@snowman.net
In reply to: Daniel Farina (#4)
Re: 9.2 Cascading replication after slave promotion

* Daniel Farina (daniel@heroku.com) wrote:

On Mon, Aug 13, 2012 at 5:32 PM, Josh Berkus <josh@agliodbs.com> wrote:

To date, I seem to be the only one convinced that this is an actual
deficiency ...

It is an actual deficiency, and I am convinced.

Yeah, I think there's more people that agree with this use-case than you
seem to think.. That said, I appreciate that it's not a trivial thing
to support cleanly.

Thanks,

Stephen

#6Josh Berkus
josh@agliodbs.com
In reply to: Stephen Frost (#5)
Re: 9.2 Cascading replication after slave promotion

Yeah, I think there's more people that agree with this use-case than you
seem to think.. That said, I appreciate that it's not a trivial thing
to support cleanly.

Not trivial, no, but not major either. Really what needs to happen is
for the timeline change record to get transmitted over the WAL stream.

Hmmm. You know, I bet I could get stream-only remastering working in an
unsafe way just by disabling the timeline checks. Time to test ...

--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com

#7Gibheer
gibheer@zero-knowledge.org
In reply to: Josh Berkus (#6)
Re: 9.2 Cascading replication after slave promotion

On Tue, 14 Aug 2012 10:50:07 -0700
Josh Berkus <josh@agliodbs.com> wrote:

Yeah, I think there's more people that agree with this use-case
than you seem to think.. That said, I appreciate that it's not a
trivial thing to support cleanly.

Not trivial, no, but not major either. Really what needs to happen is
for the timeline change record to get transmitted over the WAL stream.

Hmmm. You know, I bet I could get stream-only remastering working in
an unsafe way just by disabling the timeline checks. Time to test ...

Isn't that, what recovery_target_timeline in the recovery.conf already
does? It switches to the next timeline after a master migration. See
http://www.postgresql.org/docs/current/static/recovery-target-settings.html
for further information.