Skip to content

G
Gpdb

Greenplum / Gpdb

通知 7
Star 1
Fork 0
G
Gpdb

前往新版Gitcode，体验更适合开发者的 AI 搜索 >>

提交 1348afc0 编写于 作者: H Hao Wu 提交者： Haozhou Wang
浏览文件
操作

Fix auto explain init file 

Fix auto_explain init file for #7195

This patch includes

1. fix init_file
2. use float number for memory usage
3. Update the last SQL whose running time may be less than 1ms
4. Update test case: use newly created table other than pg_class
5. Add answer file for orca & enable nestloop
上级 73ca7e77
隐藏空白更改
内联 并排
Showing with 197 addition and 67 deletion
contrib/auto_explain/expected/auto_explain.out
浏览文件 @ 1348afc0
CREATE SCHEMA auto_explain_test;
CREATE TABLE auto_explain_test.t1(a int);
NOTICE: Table doesn't have 'DISTRIBUTED BY' clause -- Using column named 'a' as the Greenplum Database data distribution key for this table.
HINT: The 'DISTRIBUTED BY' clause determines the distribution of data. Make sure column(s) chosen are the optimal data distribution key to minimize skew.
INSERT INTO auto_explain_test.t1 VALUES(generate_series(0, 1000));
CREATE TABLE auto_explain_test.t2(b int);
NOTICE: Table doesn't have 'DISTRIBUTED BY' clause -- Using column named 'b' as the Greenplum Database data distribution key for this table.
HINT: The 'DISTRIBUTED BY' clause determines the distribution of data. Make sure column(s) chosen are the optimal data distribution key to minimize skew.
INSERT INTO auto_explain_test.t2 VALUES(generate_series(0, 1000));
SET enable_nestloop = ON;
SET CLIENT_MIN_MESSAGES = LOG;
LOAD 'auto_explain';
LOG: statement: LOAD 'auto_explain';
......@@ -17,11 +27,11 @@ SET auto_explain.log_verbose = FALSE;
LOG: statement: SET auto_explain.log_verbose = FALSE;
SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: statement: SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: duration: 0.198 ms plan:
LOG: duration: 0.173 ms plan:
Query Text: SELECT relname FROM pg_class WHERE relname='pg_class';
Seq Scan on pg_class (cost=0.00..11.46 rows=1 width=64) (actual rows=1 loops=1)
Seq Scan on pg_class (cost=0.00..12.46 rows=1 width=64) (actual rows=1 loops=1)
Filter: (relname = 'pg_class'::name)
Rows Removed by Filter: 436
Rows Removed by Filter: 438
(slice0) Executor memory: 32K bytes.
Memory used: 128000kB
relname
......@@ -29,25 +39,25 @@ Memory used: 128000kB
pg_class
(1 row)
SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
LOG: statement: SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
LOG: duration: 5.369 ms plan:
Query Text: SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
Aggregate (cost=19.62..19.64 rows=1 width=1) (actual rows=1 loops=1)
-> Hash Join (cost=4.29..19.29 rows=135 width=0) (actual rows=135 loops=1)
Hash Cond: (pg_class.oid = pg_index.indrelid)
Extra Text: Hash chain length 1.5 avg, 3 max, using 91 of 524288 buckets.
-> Seq Scan on pg_class (cost=0.00..10.37 rows=437 width=4) (actual rows=437 loops=1)
-> Hash (cost=2.60..2.60 rows=45 width=4) (actual rows=135 loops=1)
Buckets: 524288 Batches: 1 Memory Usage: 4kB
-> Seq Scan on pg_index (cost=0.00..2.60 rows=135 width=4) (actual rows=135 loops=1)
Filter: indisunique
Rows Removed by Filter: 25
(slice0) Executor memory: 4177K bytes. Work_mem: 4K bytes max.
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: statement: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: duration: 165.358 ms plan:
Query Text: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
Aggregate (cost=10000035148.67..10000035148.68 rows=1 width=8) (actual rows=1 loops=1)
-> Gather Motion 3:1 (slice2; segments: 3) (cost=10000035148.61..10000035148.66 rows=1 width=8) (actual rows=3 loops=1)
-> Aggregate (cost=10000035148.61..10000035148.62 rows=1 width=8) (actual rows=1 loops=1)
-> Nested Loop (cost=10000000000.00..10000032643.60 rows=334001 width=0) (actual rows=340340 loops=1)
-> Seq Scan on t1 (cost=0.00..13.01 rows=334 width=0) (actual rows=340 loops=1)
-> Materialize (cost=0.00..68.06 rows=1001 width=0) (actual rows=1001 loops=340)
-> Broadcast Motion 3:3 (slice1; segments: 3) (cost=0.00..53.05 rows=1001 width=0) (actual rows=1001 loops=1)
-> Seq Scan on t2 (cost=0.00..13.01 rows=334 width=0) (actual rows=340 loops=1)
(slice0) Executor memory: 131K bytes.
(slice1) Executor memory: 53K bytes avg x 3 workers, 58K bytes max (seg1).
(slice2) Executor memory: 216K bytes avg x 3 workers, 216K bytes max (seg0).
Memory used: 128000kB
?column?
----------
t
count
---------
1002001
(1 row)
SET auto_explain.log_min_duration = 1;
......@@ -56,6 +66,7 @@ SET auto_explain.log_triggers = FALSE;
LOG: statement: SET auto_explain.log_triggers = FALSE;
SET auto_explain.log_verbose = TRUE;
LOG: statement: SET auto_explain.log_verbose = TRUE;
-- this select should not dump execution plan
SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: statement: SELECT relname FROM pg_class WHERE relname='pg_class';
relname
......@@ -63,30 +74,39 @@ LOG: statement: SELECT relname FROM pg_class WHERE relname='pg_class';
pg_class
(1 row)
SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
LOG: statement: SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
LOG: duration: 4.530 ms plan:
Query Text: SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
Aggregate (cost=19.62..19.64 rows=1 width=1) (actual rows=1 loops=1)
Output: (count(*) > 0)
-> Hash Join (cost=4.29..19.29 rows=135 width=0) (actual rows=135 loops=1)
Hash Cond: (pg_class.oid = pg_index.indrelid)
Executor Memory: 4kB Segments: 1 Max: 4kB (segment -1)
work_mem: 4kB Segments: 1 Max: 4kB (segment -1) Workfile: (0 spilling)
Extra Text: Hash chain length 1.5 avg, 3 max, using 91 of 524288 buckets.
-> Seq Scan on pg_catalog.pg_class (cost=0.00..10.37 rows=437 width=4) (actual rows=437 loops=1)
Output: pg_class.oid
-> Hash (cost=2.60..2.60 rows=45 width=4) (actual rows=135 loops=1)
Output: pg_index.indrelid
Buckets: 524288 Batches: 1 Memory Usage: 4kB
-> Seq Scan on pg_catalog.pg_index (cost=0.00..2.60 rows=135 width=4) (actual rows=135 loops=1)
Output: pg_index.indrelid
Filter: pg_index.indisunique
Rows Removed by Filter: 25
(slice0) Executor memory: 4177K bytes. Work_mem: 4K bytes max.
-- this select should also dump plan, since it takes too much time to run
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: statement: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: duration: 208.523 ms plan:
Query Text: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
Aggregate (cost=10000035148.67..10000035148.68 rows=1 width=8) (actual rows=1 loops=1)
Output: count((count(*)))
-> Gather Motion 3:1 (slice2; segments: 3) (cost=10000035148.61..10000035148.66 rows=1 width=8) (actual rows=3 loops=1)
Output: (count(*))
-> Aggregate (cost=10000035148.61..10000035148.62 rows=1 width=8) (actual rows=1 loops=1)
Output: count(*)
-> Nested Loop (cost=10000000000.00..10000032643.60 rows=334001 width=0) (actual rows=340340 loops=1)
-> Seq Scan on auto_explain_test.t1 (cost=0.00..13.01 rows=334 width=0) (actual rows=340 loops=1)
Output: t1.a
-> Materialize (cost=0.00..68.06 rows=1001 width=0) (actual rows=1001 loops=340)
Output: t2.b
-> Broadcast Motion 3:3 (slice1; segments: 3) (cost=0.00..53.05 rows=1001 width=0) (actual rows=1001 loops=1)
Output: t2.b
-> Seq Scan on auto_explain_test.t2 (cost=0.00..13.01 rows=334 width=0) (actual rows=340 loops=1)
Output: t2.b
(slice0) Executor memory: 131K bytes.
(slice1) Executor memory: 43K bytes avg x 3 workers, 43K bytes max (seg0).
(slice2) Executor memory: 216K bytes avg x 3 workers, 216K bytes max (seg0).
Memory used: 128000kB
?column?
----------
t
count
---------
1002001
(1 row)
-- clean jobs
DROP TABLE auto_explain_test.t1;
LOG: statement: DROP TABLE auto_explain_test.t1;
DROP TABLE auto_explain_test.t2;
LOG: statement: DROP TABLE auto_explain_test.t2;
DROP SCHEMA auto_explain_test;
LOG: statement: DROP SCHEMA auto_explain_test;
contrib/auto_explain/expected/auto_explain_optimizer.out 0 → 100644
浏览文件 @ 1348afc0
CREATE SCHEMA auto_explain_test;
CREATE TABLE auto_explain_test.t1(a int);
NOTICE: Table doesn't have 'DISTRIBUTED BY' clause -- Using column named 'a' as the Greenplum Database data distribution key for this table.
HINT: The 'DISTRIBUTED BY' clause determines the distribution of data. Make sure column(s) chosen are the optimal data distribution key to minimize skew.
INSERT INTO auto_explain_test.t1 VALUES(generate_series(0, 1000));
CREATE TABLE auto_explain_test.t2(b int);
NOTICE: Table doesn't have 'DISTRIBUTED BY' clause -- Using column named 'b' as the Greenplum Database data distribution key for this table.
HINT: The 'DISTRIBUTED BY' clause determines the distribution of data. Make sure column(s) chosen are the optimal data distribution key to minimize skew.
INSERT INTO auto_explain_test.t2 VALUES(generate_series(0, 1000));
SET enable_nestloop = ON;
SET CLIENT_MIN_MESSAGES = LOG;
LOAD 'auto_explain';
LOG: statement: LOAD 'auto_explain';
SET auto_explain.log_analyze = TRUE;
LOG: statement: SET auto_explain.log_analyze = TRUE;
SET auto_explain.log_min_duration = 0;
LOG: statement: SET auto_explain.log_min_duration = 0;
SET auto_explain.log_buffers = FALSE;
LOG: statement: SET auto_explain.log_buffers = FALSE;
SET auto_explain.log_triggers = TRUE;
LOG: statement: SET auto_explain.log_triggers = TRUE;
SET auto_explain.log_nested_statements = FALSE;
LOG: statement: SET auto_explain.log_nested_statements = FALSE;
SET auto_explain.log_timing = FALSE;
LOG: statement: SET auto_explain.log_timing = FALSE;
SET auto_explain.log_verbose = FALSE;
LOG: statement: SET auto_explain.log_verbose = FALSE;
SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: statement: SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: Planner produced plan :0
LOG: duration: 0.112 ms plan:
Query Text: SELECT relname FROM pg_class WHERE relname='pg_class';
Seq Scan on pg_class (cost=0.00..12.46 rows=1 width=64) (actual rows=1 loops=1)
Filter: (relname = 'pg_class'::name)
Rows Removed by Filter: 438
(slice0) Executor memory: 32K bytes.
Memory used: 128000kB
relname
----------
pg_class
(1 row)
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: statement: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: duration: 195.785 ms plan:
Query Text: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
Aggregate (cost=0.00..1326086.34 rows=1 width=8) (actual rows=1 loops=1)
-> Gather Motion 3:1 (slice2; segments: 3) (cost=0.00..1326086.34 rows=1 width=8) (actual rows=3 loops=1)
-> Aggregate (cost=0.00..1326086.34 rows=1 width=8) (actual rows=1 loops=1)
-> Nested Loop (cost=0.00..1326086.34 rows=334001 width=1) (actual rows=340340 loops=1)
Join Filter: true
-> Broadcast Motion 3:3 (slice1; segments: 3) (cost=0.00..431.02 rows=1001 width=1) (actual rows=1001 loops=1)
-> Seq Scan on t1 (cost=0.00..431.01 rows=334 width=1) (actual rows=340 loops=1)
-> Seq Scan on t2 (cost=0.00..431.01 rows=334 width=1) (actual rows=340 loops=1002)
(slice0) Executor memory: 67K bytes.
(slice1) Executor memory: 42K bytes avg x 3 workers, 42K bytes max (seg0).
(slice2) Executor memory: 119K bytes avg x 3 workers, 119K bytes max (seg0).
Memory used: 128000kB
count
---------
1002001
(1 row)
SET auto_explain.log_min_duration = 1;
LOG: statement: SET auto_explain.log_min_duration = 1;
SET auto_explain.log_triggers = FALSE;
LOG: statement: SET auto_explain.log_triggers = FALSE;
SET auto_explain.log_verbose = TRUE;
LOG: statement: SET auto_explain.log_verbose = TRUE;
-- this select should not dump execution plan
SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: statement: SELECT relname FROM pg_class WHERE relname='pg_class';
LOG: Planner produced plan :0
relname
----------
pg_class
(1 row)
-- this select should also dump plan, since it takes too much time to run
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: statement: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
LOG: duration: 180.126 ms plan:
Query Text: SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
Aggregate (cost=0.00..1326086.34 rows=1 width=8) (actual rows=1 loops=1)
Output: count((count()))
-> Gather Motion 3:1 (slice2; segments: 3) (cost=0.00..1326086.34 rows=1 width=8) (actual rows=3 loops=1)
Output: (count())
-> Aggregate (cost=0.00..1326086.34 rows=1 width=8) (actual rows=1 loops=1)
Output: count()
-> Nested Loop (cost=0.00..1326086.34 rows=334001 width=1) (actual rows=340340 loops=1)
Join Filter: true
-> Broadcast Motion 3:3 (slice1; segments: 3) (cost=0.00..431.02 rows=1001 width=1) (actual rows=1001 loops=1)
-> Seq Scan on auto_explain_test.t1 (cost=0.00..431.01 rows=334 width=1) (actual rows=340 loops=1)
-> Seq Scan on auto_explain_test.t2 (cost=0.00..431.01 rows=334 width=1) (actual rows=340 loops=1002)
(slice0) Executor memory: 67K bytes.
(slice1) Executor memory: 42K bytes avg x 3 workers, 42K bytes max (seg0).
(slice2) Executor memory: 119K bytes avg x 3 workers, 119K bytes max (seg0).
Memory used: 128000kB
count
---------
1002001
(1 row)
-- clean jobs
DROP TABLE auto_explain_test.t1;
LOG: statement: DROP TABLE auto_explain_test.t1;
DROP TABLE auto_explain_test.t2;
LOG: statement: DROP TABLE auto_explain_test.t2;
DROP SCHEMA auto_explain_test;
LOG: statement: DROP SCHEMA auto_explain_test;
contrib/auto_explain/init_file
浏览文件 @ 1348afc0
-- start_matchsubs
m/duration: .*$/
s/duration:\s+[0-9.]+\s*ms.*/duration: ms/
m/\(cost=[a-z_A-Z0-9. =]+\)\s+\([a-z_A-Z0-9. =]+\)/
s/\(cost=[a-z_A-Z0-9. =]+\)\s+\([a-z_A-Z0-9. =]+\)//
m/duration:.*$/
s/duration:\s+[0-9.]+\s*ms.*/duration: ms/
m/Rows Removed by Filter: .*/
s/Rows Removed by Filter: .*/Rows Removed by Filter: /
m/Executor [mM]emory: .*/
s/Executor [mM]emory: .*/Executor Memory: /
m/Memory used:\s+[0-9kKmMgGB]+/
s/Memory used:\s+[0-9kKmMgGB]+/Memory used:/
m/Buckets: [0-9]+\s+Batches: [0-9]+\s+Memory Usage:\s+[0-9kKmMgG]B/
s/Buckets: [0-9]+\s+Batches: [0-9]+\s+Memory Usage:\s+[0-9kKmMgG]B/Buckets: 524288 Batches: 1 Memory Usage: 4kB/
m/Rows Removed by Filter: .*/
s/Rows Removed by Filter: .*/Rows Removed by Filter: /
m/Extra Text: Hash chain length [0-9.]+ avg, [0-9]+ max, using [0-9]+ of [0-9]+ buckets/
s/Extra Text: Hash chain length [0-9.]+ avg, [0-9]+ max, using [0-9]+ of [0-9]+ buckets/Extra Text: Hash chain length 1.5 avg, 3 max, using 91 of 524288 buckets/
m/Memory used:\s+[0-9]+kB.*/
s/Memory used:\s+[0-9]+kB.*/Memory used:/
m/work_mem: [0-9kKmMgG]B Segments: [-0-9]+ Max: [0-9kKmMgG]+B.*/
s/work_mem: [0-9kKmMgG]B Segments: [-0-9]+ Max: [0-9kKmMgG]+B.*/work_mem: 4kB Segments: 1 Max: 4kB/
-- end_matchsubs
-- start_matchignore
# orca failed to produce the expected plan, ignore this message
m/LOG: Planner produced plan :0/
-- end_matchignore
contrib/auto_explain/sql/auto_explain.sql
浏览文件 @ 1348afc0
CREATE SCHEMA auto_explain_test;
CREATE TABLE auto_explain_test.t1(a int);
INSERT INTO auto_explain_test.t1 VALUES(generate_series(0, 1000));
CREATE TABLE auto_explain_test.t2(b int);
INSERT INTO auto_explain_test.t2 VALUES(generate_series(0, 1000));
SET enable_nestloop = ON;
SET CLIENT_MIN_MESSAGES = LOG;
LOAD 'auto_explain';
SET auto_explain.log_analyze = TRUE;
......@@ -9,11 +16,18 @@ SET auto_explain.log_timing = FALSE;
SET auto_explain.log_verbose = FALSE;
SELECT relname FROM pg_class WHERE relname='pg_class';
SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
SET auto_explain.log_min_duration = 1;
SET auto_explain.log_triggers = FALSE;
SET auto_explain.log_verbose = TRUE;
-- this select should not dump execution plan
SELECT relname FROM pg_class WHERE relname='pg_class';
SELECT count(*)>0 FROM pg_class, pg_index WHERE oid = indrelid AND indisunique;
-- this select should also dump plan, since it takes too much time to run
SELECT count(*) FROM auto_explain_test.t1, auto_explain_test.t2;
-- clean jobs
DROP TABLE auto_explain_test.t1;
DROP TABLE auto_explain_test.t2;
DROP SCHEMA auto_explain_test;
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或