Horton

Re: Minimal logical decoding on standbys

Jeff Davis <pgsql@j-davis.com>

View thread
From: Jeff Davis <pgsql@j-davis.com>
To: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>, Andres Freund <andres@anarazel.de>
Cc: Robert Haas <robertmhaas@gmail.com>, Thomas Munro <thomas.munro@gmail.com>, Alvaro Herrera <alvherre@2ndquadrant.com>, Ibrar Ahmed <ibrar.ahmad@gmail.com>, Amit Khandekar <amitdkhan.pg@gmail.com>, fabriziomello@gmail.com, tushar <tushar.ahuja@enterprisedb.com>, Rahila Syed <rahila.syed@2ndquadrant.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2023-03-31T09:44:33Z
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.

Attachments

On Wed, 2023-03-08 at 11:25 +0100, Drouvot, Bertrand wrote:
> - Maybe ConditionVariableEventSleep() should take care of the
> “WaitEventSetWait returns 1 and cvEvent.event == WL_POSTMASTER_DEATH”
> case?

Thank you, done. I think the nearby line was also wrong, returning true
when there was no timeout. I combined the lines and got rid of the
early return so it can check the list and timeout condition like
normal. Attached.

> - Maybe ConditionVariableEventSleep() could accept and deal with the
> CV being NULL?
> I used it in the POC attached to handle logical decoding on the
> primary server case.
> One option should be to create a dedicated CV for that case though.

I don't think it's a good idea to have a CV-based API that doesn't need
a CV. Wouldn't that just be a normal WaitEventSet?

> - In the POC attached I had to add this extra condition “(cv &&
> !RecoveryInProgress())” to avoid waiting on the timeout when there is
> a promotion.
> That makes me think that we may want to add 2 extra parameters (as 2
> functions returning a bool?) to ConditionVariableEventSleep()
> to check whether or not we still want to test the socket or the CV
> wake up in each loop iteration.

That seems like a complex API. Would it work to signal the CV during
promotion instead?

> Also 3 additional remarks:
> 
> 1) About InitializeConditionVariableWaitSet() and
> ConditionVariableWaitSetCreate(): I'm not sure about the naming as
> there is no CV yet (they "just" deal with WaitEventSet).

It's a WaitEventSet that contains the conditions always required for
any CV, and allows you to add in more.

> 3)
> 
> I wonder if there is no race conditions: ConditionVariableWaitSet is
> being initialized with PGINVALID_SOCKET
> as WL_LATCH_SET and might be also (if IsUnderPostmaster) be
> initialized with PGINVALID_SOCKET as WL_EXIT_ON_PM_DEATH.
> 
> So IIUC, the patch is introducing 2 new possible source of wake up.

Those should be the same conditions already required by
ConditionVariableTimedSleep() in master, right?


Regards,
	Jeff Davis