Skip to content

Commit a592ed9

Browse files
author
Amit Kapila
committed
Fix assertion failures while processing NEW_CID record in logical decoding.
When the logical decoding restarts from NEW_CID, since there is no association between the top transaction and its subtransaction, both are created as top transactions and have the same LSN. This caused the assertion failure in AssertTXNLsnOrder(). This patch skips the assertion check until we reach the LSN at which we start decoding the contents of the transaction, specifically start_decoding_at LSN in SnapBuild. This is okay because we don't guarantee to make the association between top transaction and subtransaction until we try to decode the actual contents of transaction. The ordering of the records prior to the start_decoding_at LSN should have been checked before the restart. The other assertion failure is due to the reason that we forgot to track that we have considered top-level transaction id in the list of catalog changing transactions that were committed when one of its subtransactions is marked as containing catalog change. Reported-by: Tomas Vondra, Osumi Takamichi Author: Masahiko Sawada, Kuroda Hayato Reviewed-by: Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi, Masahiko Sawada Backpatch-through: 10 Discussion: https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.com Discussion: https://postgr.es/m/TYCPR01MB83733C6CEAE47D0280814D5AED7A9%40TYCPR01MB8373.jpnprd01.prod.outlook.com
1 parent d033f8f commit a592ed9

File tree

4 files changed

+78
-0
lines changed

4 files changed

+78
-0
lines changed

‎contrib/test_decoding/expected/catalog_change_snapshot.out

Lines changed: 45 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -42,3 +42,48 @@ COMMIT
4242
stop
4343
(1 row)
4444

45+
46+
starting permutation: s0_init s0_begin s0_savepoint s0_insert s1_checkpoint s1_get_changes s0_insert2 s0_commit s0_begin s0_insert s1_checkpoint s1_get_changes s0_commit s1_get_changes
47+
step s0_init: SELECT 'init' FROM pg_create_logical_replication_slot('isolation_slot', 'test_decoding');
48+
?column?
49+
--------
50+
init
51+
(1 row)
52+
53+
step s0_begin: BEGIN;
54+
step s0_savepoint: SAVEPOINT sp1;
55+
step s0_insert: INSERT INTO tbl1 VALUES (1);
56+
step s1_checkpoint: CHECKPOINT;
57+
step s1_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'skip-empty-xacts', '1', 'include-xids', '0');
58+
data
59+
----
60+
(0 rows)
61+
62+
step s0_insert2: INSERT INTO user_cat VALUES (1);
63+
step s0_commit: COMMIT;
64+
step s0_begin: BEGIN;
65+
step s0_insert: INSERT INTO tbl1 VALUES (1);
66+
step s1_checkpoint: CHECKPOINT;
67+
step s1_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'skip-empty-xacts', '1', 'include-xids', '0');
68+
data
69+
-------------------------------------------------------------
70+
BEGIN
71+
table public.tbl1: INSERT: val1[integer]:1 val2[integer]:null
72+
table public.user_cat: INSERT: val1[integer]:1
73+
COMMIT
74+
(4 rows)
75+
76+
step s0_commit: COMMIT;
77+
step s1_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'skip-empty-xacts', '1', 'include-xids', '0');
78+
data
79+
-------------------------------------------------------------
80+
BEGIN
81+
table public.tbl1: INSERT: val1[integer]:1 val2[integer]:null
82+
COMMIT
83+
(3 rows)
84+
85+
?column?
86+
--------
87+
stop
88+
(1 row)
89+

‎contrib/test_decoding/specs/catalog_change_snapshot.spec

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -4,11 +4,13 @@ setup
44
{
55
DROP TABLE IF EXISTS tbl1;
66
CREATE TABLE tbl1 (val1 integer, val2 integer);
7+
CREATE TABLE user_cat (val1 integer) WITH (user_catalog_table = true);
78
}
89

910
teardown
1011
{
1112
DROP TABLE tbl1;
13+
DROP TABLE user_cat;
1214
SELECT 'stop' FROM pg_drop_replication_slot('isolation_slot');
1315
}
1416

@@ -19,6 +21,7 @@ step "s0_begin" { BEGIN; }
1921
step "s0_savepoint" { SAVEPOINT sp1; }
2022
step "s0_truncate" { TRUNCATE tbl1; }
2123
step "s0_insert" { INSERT INTO tbl1 VALUES (1); }
24+
step "s0_insert2" { INSERT INTO user_cat VALUES (1); }
2225
step "s0_commit" { COMMIT; }
2326

2427
session "s1"
@@ -37,3 +40,16 @@ step "s1_get_changes" { SELECT data FROM pg_logical_slot_get_changes('isolation_
3740
# record written by bgwriter. One might think we can either stop the bgwriter or
3841
# increase LOG_SNAPSHOT_INTERVAL_MS but it's not practical via tests.
3942
permutation "s0_init" "s0_begin" "s0_savepoint" "s0_truncate" "s1_checkpoint" "s1_get_changes" "s0_commit" "s0_begin" "s0_insert" "s1_checkpoint" "s1_get_changes" "s0_commit" "s1_get_changes"
43+
44+
# Test that we can handle the case where there is no association between top-level
45+
# transaction and its subtransactions. The last decoding restarts from the first
46+
# checkpoint, decodes NEW_CID generated by "s0_insert2", and marks the subtransaction
47+
# as containing catalog changes while adding tuple cids to its top-level transaction.
48+
# During that, both transaction entries are created in ReorderBuffer as top-level
49+
# transactions and have the same LSN. We check if the assertion check for the order
50+
# of transaction LSNs in AssertTXNLsnOrder() is skipped since we are still before the
51+
# LSN at which we start replaying the contents of transactions. Besides, when decoding
52+
# the commit record of the top-level transaction, we must force the top-level
53+
# transaction to do timetravel since one of its subtransactions has been marked as
54+
# containing catalog changes.
55+
permutation "s0_init" "s0_begin" "s0_savepoint" "s0_insert" "s1_checkpoint" "s1_get_changes" "s0_insert2" "s0_commit" "s0_begin" "s0_insert" "s1_checkpoint" "s1_get_changes" "s0_commit" "s1_get_changes"

‎src/backend/replication/logical/reorderbuffer.c

Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -871,10 +871,24 @@ static void
871871
AssertTXNLsnOrder(ReorderBuffer *rb)
872872
{
873873
#ifdef USE_ASSERT_CHECKING
874+
LogicalDecodingContext *ctx = rb->private_data;
874875
dlist_iter iter;
875876
XLogRecPtr prev_first_lsn = InvalidXLogRecPtr;
876877
XLogRecPtr prev_base_snap_lsn = InvalidXLogRecPtr;
877878

879+
/*
880+
* Skip the verification if we don't reach the LSN at which we start
881+
* decoding the contents of transactions yet because until we reach the
882+
* LSN, we could have transactions that don't have the association between
883+
* the top-level transaction and subtransaction yet and consequently have
884+
* the same LSN. We don't guarantee this association until we try to
885+
* decode the actual contents of transaction. The ordering of the records
886+
* prior to the start_decoding_at LSN should have been checked before the
887+
* restart.
888+
*/
889+
if (SnapBuildXactNeedsSkip(ctx->snapshot_builder, ctx->reader->EndRecPtr))
890+
return;
891+
878892
dlist_foreach(iter, &rb->toplevel_by_lsn)
879893
{
880894
ReorderBufferTXN *cur_txn = dlist_container(ReorderBufferTXN, node,

‎src/backend/replication/logical/snapbuild.c

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1092,6 +1092,9 @@ SnapBuildCommitTxn(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid,
10921092
else if (sub_needs_timetravel)
10931093
{
10941094
/* track toplevel txn as well, subxact alone isn't meaningful */
1095+
elog(DEBUG2, "forced transaction %u to do timetravel due to one of its subtransactions",
1096+
xid);
1097+
needs_timetravel = true;
10951098
SnapBuildAddCommittedTxn(builder, xid);
10961099
}
10971100
else if (needs_timetravel)

0 commit comments

Comments
 (0)