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

Commit8d1c1ca

Browse files
committed
Reduce cost of test_decoding's new oldest_xmin test
Change a whole-database VACUUM into doing just pg_attribute, which isthe portion that verifies what we want it to do. The originalformulation wastes a lot of CPU time, which leads the test to fail whenruntime exceeds isolationtester timeout when it's super-slow, such asunder CLOBBER_CACHE_ALWAYS. Per buildfarm member friarbird.It turns out that the previous shape of the test doesn't always detectthe condition it is supposed to detect (on unpatched reorderbuffercode): the reason is that there is a good chance of encountering axl_running_xacts record (logged every 15 seconds) before the checkpoint-- and because we advance the xmin when we receive that WAL record, andwe *don't* advance the xmin twice consecutively without receiving aclient message in between, that means the xmin is not advanced enoughfor the tuple to be pruned from pg_attribute by VACUUM. So the testwould spuriously pass.The reason this test deficiency wasn't detected earlier is that HOTpruning removes the tuple anyway, even if vacuum leaves it in place, sothe test correctly fails (detecting the coding mistake), but for thewrong reason.To fix this mess, run the s0_get_changes step twice before vacuuminstead of once: this seems to cause the xmin to be advanced reliably,wreaking havoc with more certainty.Author: Arseny SherDiscussion:https://postgr.es/m/87h8lkuxoa.fsf@ars-thinkpad
1 parentf61988d commit8d1c1ca

File tree

2 files changed

+11
-5
lines changed

2 files changed

+11
-5
lines changed

‎contrib/test_decoding/expected/oldest_xmin.out

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
Parsed test spec with 2 sessions
22

3-
starting permutation: s0_begin s0_getxid s1_begin s1_insert s0_alter s0_commit s0_checkpoint s0_get_changes s1_commit s0_vacuum s0_get_changes
3+
starting permutation: s0_begin s0_getxid s1_begin s1_insert s0_alter s0_commit s0_checkpoint s0_get_changess0_get_changess1_commit s0_vacuum s0_get_changes
44
step s0_begin: BEGIN;
55
step s0_getxid: SELECT txid_current() IS NULL;
66
?column?
@@ -14,8 +14,11 @@ step s0_checkpoint: CHECKPOINT;
1414
step s0_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'include-xids', '0', 'skip-empty-xacts', '1');
1515
data
1616

17+
step s0_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'include-xids', '0', 'skip-empty-xacts', '1');
18+
data
19+
1720
step s1_commit: COMMIT;
18-
step s0_vacuum: VACUUMFULL;
21+
step s0_vacuum: VACUUMpg_attribute;
1922
step s0_get_changes: SELECT data FROM pg_logical_slot_get_changes('isolation_slot', NULL, NULL, 'include-xids', '0', 'skip-empty-xacts', '1');
2023
data
2124

‎contrib/test_decoding/specs/oldest_xmin.spec

Lines changed: 6 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -22,16 +22,19 @@ step "s0_getxid" { SELECT txid_current() IS NULL; }
2222
step"s0_alter" {ALTERTYPEbasketDROPATTRIBUTEmangos; }
2323
step"s0_commit" {COMMIT; }
2424
step"s0_checkpoint" {CHECKPOINT; }
25-
step"s0_vacuum" {VACUUMFULL; }
25+
step"s0_vacuum" {VACUUMpg_attribute; }
2626
step"s0_get_changes" {SELECTdataFROMpg_logical_slot_get_changes('isolation_slot',NULL,NULL,'include-xids','0','skip-empty-xacts','1'); }
2727

2828
session"s1"
2929
step"s1_begin" {BEGIN; }
3030
step"s1_insert" {INSERTINTOharvestVALUES ((1,2,3)); }
3131
step"s1_commit" {COMMIT; }
3232

33-
# Checkpoint with following get_changes forces to advance xmin. ALTER of a
33+
# Checkpoint with following get_changes forces xmin advancement. We do
34+
# get_changes twice because if one more xl_running_xacts record had slipped
35+
# before our CHECKPOINT, xmin will be advanced only on this record, thus not
36+
# reaching value needed for vacuuming corresponding pg_attribute entry. ALTER of
3437
# composite type is a rare form of DDL which allows T1 to see the tuple which
3538
# will be removed (xmax set) before T1 commits. That is, interlocking doesn't
3639
# forbid modifying catalog after someone read it (and didn't commit yet).
37-
permutation"s0_begin""s0_getxid""s1_begin""s1_insert""s0_alter""s0_commit""s0_checkpoint""s0_get_changes""s1_commit""s0_vacuum""s0_get_changes"
40+
permutation"s0_begin""s0_getxid""s1_begin""s1_insert""s0_alter""s0_commit""s0_checkpoint""s0_get_changes""s0_get_changes""s1_commit""s0_vacuum""s0_get_changes"

0 commit comments

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