Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit64ff0fe

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 noassociation between the top transaction and its subtransaction, both arecreated as top transactions and have the same LSN. This caused theassertion failure in AssertTXNLsnOrder().This patch skips the assertion check until we reach the LSN at which westart decoding the contents of the transaction, specificallystart_decoding_at LSN in SnapBuild. This is okay because we don'tguarantee to make the association between top transaction andsubtransaction until we try to decode the actual contents of transaction.The ordering of the records prior to the start_decoding_at LSN should havebeen checked before the restart.The other assertion failure is due to the reason that we forgot to trackthat we have considered top-level transaction id in the list of catalogchanging transactions that were committed when one of its subtransactionsis marked as containing catalog change.Reported-by: Tomas Vondra, Osumi TakamichiAuthor: Masahiko Sawada, Kuroda HayatoReviewed-by: Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi, Masahiko SawadaBackpatch-through: 10Discussion:https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.comDiscussion:https://postgr.es/m/TYCPR01MB83733C6CEAE47D0280814D5AED7A9%40TYCPR01MB8373.jpnprd01.prod.outlook.com
1 parentaf64846 commit64ff0fe

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
DROPTABLEIFEXISTStbl1;
66
CREATETABLEtbl1(val1integer,val2integer);
7+
CREATETABLEuser_cat(val1integer)WITH(user_catalog_table=true);
78
}
89

910
teardown
1011
{
1112
DROPTABLEtbl1;
13+
DROPTABLEuser_cat;
1214
SELECT'stop'FROMpg_drop_replication_slot('isolation_slot');
1315
}
1416

@@ -19,6 +21,7 @@ step "s0_begin" { BEGIN; }
1921
step"s0_savepoint"{SAVEPOINTsp1;}
2022
step"s0_truncate"{TRUNCATEtbl1;}
2123
step"s0_insert"{INSERTINTOtbl1VALUES(1);}
24+
step"s0_insert2"{INSERTINTOuser_catVALUES(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
@@ -878,10 +878,24 @@ static void
878878
AssertTXNLsnOrder(ReorderBuffer*rb)
879879
{
880880
#ifdefUSE_ASSERT_CHECKING
881+
LogicalDecodingContext*ctx=rb->private_data;
881882
dlist_iteriter;
882883
XLogRecPtrprev_first_lsn=InvalidXLogRecPtr;
883884
XLogRecPtrprev_base_snap_lsn=InvalidXLogRecPtr;
884885

886+
/*
887+
* Skip the verification if we don't reach the LSN at which we start
888+
* decoding the contents of transactions yet because until we reach the
889+
* LSN, we could have transactions that don't have the association between
890+
* the top-level transaction and subtransaction yet and consequently have
891+
* the same LSN. We don't guarantee this association until we try to
892+
* decode the actual contents of transaction. The ordering of the records
893+
* prior to the start_decoding_at LSN should have been checked before the
894+
* restart.
895+
*/
896+
if (SnapBuildXactNeedsSkip(ctx->snapshot_builder,ctx->reader->EndRecPtr))
897+
return;
898+
885899
dlist_foreach(iter,&rb->toplevel_by_lsn)
886900
{
887901
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
@@ -1101,6 +1101,9 @@ SnapBuildCommitTxn(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid,
11011101
elseif (sub_needs_timetravel)
11021102
{
11031103
/* track toplevel txn as well, subxact alone isn't meaningful */
1104+
elog(DEBUG2,"forced transaction %u to do timetravel due to one of its subtransactions",
1105+
xid);
1106+
needs_timetravel= true;
11041107
SnapBuildAddCommittedTxn(builder,xid);
11051108
}
11061109
elseif (needs_timetravel)

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp