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: 2019-05-27T11:34:44Z
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 Fri, 24 May 2019 at 21:00, Amit Khandekar <amitdkhan.pg@gmail.com> wrote:
>
> On Fri, 24 May 2019 at 19:26, Amit Khandekar <amitdkhan.pg@gmail.com> wrote:
> > Working on the patch now ....
>
> Attached is an incremental WIP patch
> handle_wal_level_changes_WIP.patch to be applied over the earlier main
> patch logical-decoding-on-standby_v4_rebased.patch.

I found an issue with these changes : When we change master wal_level
from logical to hot_standby, and again back to logical, and then
create a logical replication slot on slave, it gets created; but when
I do pg_logical_slot_get_changes() with that slot, it seems to read
records *before* I created the logical slot, so it encounters
parameter-change(logical=>hot_standby) record, so returns an error as
per the patch, because now in DecodeXLogOp() I error out when
XLOG_PARAMETER_CHANGE is found :

@@ -190,11 +190,23 @@ DecodeXLogOp(LogicalDecodingContext *ctx,
XLogRecordBuffer *buf)
             * can restart from there.
             */
            break;
+         case XLOG_PARAMETER_CHANGE:
+         {
+           xl_parameter_change *xlrec =
+             (xl_parameter_change *) XLogRecGetData(buf->record);
+
+           /* Cannot proceed if master itself does not have logical data */
+           if (xlrec->wal_level < WAL_LEVEL_LOGICAL)
+             ereport(ERROR,
+                 (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
+                  errmsg("logical decoding on standby requires "
+                     "wal_level >= logical on master")));
+           break;
+         }

I thought it won't read records *before* the slot was created. Am I
missing something ?

>
> > >
> > > On 2019-05-21 09:19:37 -0700, Andres Freund wrote:
> > > > So I suspect we need conflict handling in xlog_redo's
> > > > XLOG_PARAMETER_CHANGE case. If we there check against existing logical
> > > > slots, we ought to be safe.
>
> Yet to do this. Andres, how do you want to handle this scenario ? Just
> drop all the existing logical slots like what we decided for conflict
> recovery for conflicting xids ?

I went ahead and added handling that drops existing slots when we
encounter XLOG_PARAMETER_CHANGE in xlog_redo().

Attached is logical-decoding-on-standby_v5.patch, that contains all
the changes so far.

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