1
0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-07-29 08:21:15 +03:00

MCOL-4957 Fix performance slowdown for processing TIMESTAMP columns.

Part 1:
 As part of MCOL-3776 to address synchronization issue while accessing
 the fTimeZone member of the Func class, mutex locks were added to the
 accessor and mutator methods. However, this slows down processing
 of TIMESTAMP columns in PrimProc significantly as all threads across
 all concurrently running queries would serialize on the mutex. This
 is because PrimProc only has a single global object for the functor
 class (class derived from Func in utils/funcexp/functor.h) for a given
 function name. To fix this problem:

   (1) We remove the fTimeZone as a member of the Func derived classes
   (hence removing the mutexes) and instead use the fOperationType
   member of the FunctionColumn class to propagate the timezone values
   down to the individual functor processing functions such as
   FunctionColumn::getStrVal(), FunctionColumn::getIntVal(), etc.

   (2) To achieve (1), a timezone member is added to the
   execplan::CalpontSystemCatalog::ColType class.

Part 2:
 Several functors in the Funcexp code call dataconvert::gmtSecToMySQLTime()
 and dataconvert::mySQLTimeToGmtSec() functions for conversion between seconds
 since unix epoch and broken-down representation. These functions in turn call
 the C library function localtime_r() which currently has a known bug of holding
 a global lock via a call to __tz_convert. This significantly reduces performance
 in multi-threaded applications where multiple threads concurrently call
 localtime_r(). More details on the bug:
   https://sourceware.org/bugzilla/show_bug.cgi?id=16145

 This bug in localtime_r() caused processing of the Functors in PrimProc to
 slowdown significantly since a query execution causes Functors code to be
 processed in a multi-threaded manner.

 As a fix, we remove the calls to localtime_r() from gmtSecToMySQLTime()
 and mySQLTimeToGmtSec() by performing the timezone-to-offset conversion
 (done in dataconvert::timeZoneToOffset()) during the execution plan
 creation in the plugin. Note that localtime_r() is only called when the
 time_zone system variable is set to "SYSTEM".

 This fix also required changing the timezone type from a std::string to
 a long across the system.
This commit is contained in:
Gagan Goel
2022-02-09 19:03:00 -05:00
parent f67a37bcae
commit 973e5024d8
120 changed files with 1022 additions and 695 deletions

View File

@ -0,0 +1,141 @@
#
# Test cases for the TIMESTAMP datatype
#
DROP DATABASE IF EXISTS timestamp_test;
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a timestamp);
SET time_zone='-5:00';
INSERT INTO ctimestamp VALUES ('2019-01-01 01:02:03'), ('2019-05-05 01:01:01');
SET time_zone='+1:00';
SELECT a FROM ctimestamp ORDER BY a;
a
2019-01-01 07:02:03
2019-05-05 07:01:01
SET time_zone='-2:00';
SELECT a FROM ctimestamp ORDER BY a;
a
2019-01-01 04:02:03
2019-05-05 04:01:01
CREATE TABLE ctimestamp2 (a timestamp DEFAULT 0);
INSERT INTO ctimestamp2 SELECT * FROM ctimestamp;
SELECT a FROM ctimestamp2 ORDER BY a;
a
2019-01-01 04:02:03
2019-05-05 04:01:01
CREATE TABLE ctimestamp3 (a timestamp);
INSERT INTO ctimestamp3 VALUES (19940101), (940101),
(19940101010203), (940101010203), ('1994-01-01T01:02:03');
SELECT a FROM ctimestamp3 ORDER BY a;
a
1994-01-01 00:00:00
1994-01-01 00:00:00
1994-01-01 01:02:03
1994-01-01 01:02:03
1994-01-01 01:02:03
CREATE TABLE ctimestamp4 (a timestamp(6) default 0);
INSERT INTO ctimestamp4 VALUES (0), ('2019-01-01 01:01:01.123456');
SELECT a, microsecond(a) FROM ctimestamp4 ORDER BY a;
a microsecond(a)
0000-00-00 00:00:00.000000 0
2019-01-01 01:01:01.123456 123456
DROP DATABASE IF EXISTS timestamp_test;
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a timestamp);
SET time_zone='+0:00';
INSERT INTO ctimestamp VALUES ('2019-01-02 00:02:03'),
('2019-01-02 01:02:03'), ('2019-01-02 10:11:12');
SET time_zone='+1:00';
SELECT a, a BETWEEN '2019-01-02 02:00:00' AND '2019-01-02 13:00:00'
FROM ctimestamp ORDER BY a;
a a BETWEEN '2019-01-02 02:00:00' AND '2019-01-02 13:00:00'
2019-01-02 01:02:03 0
2019-01-02 02:02:03 1
2019-01-02 11:11:12 1
SELECT a, IF(a < '2019-01-02 02:00:00', 'yes', 'no'),
ADDTIME(a, '1:1:1'), STR_TO_DATE(a, '%Y-%m-%d %H:%i:%s'),
EXTRACT(DAY_HOUR FROM a), EXTRACT(MINUTE_SECOND FROM a),
TIME_FORMAT(a, '%H:\%i:\%s'), a RLIKE '02:03', IFNULL(NULL, a),
CASE a WHEN '2019-01-02 01:02:03' THEN 'found' WHEN '2019-01-02 11:11:12'
THEN 'found2' ELSE 'notfound' END, CHAR_LENGTH(a),
CAST(a AS UNSIGNED INT), CAST(a AS CHAR), CAST(a AS DATE),
TIME(CAST(a AS DATETIME)), TIME(COALESCE(NULL, a)), HEX(a),
NULLIF(a, '2019-01-02 01:02:03'), TIMEDIFF(a, '2019-01-01 01:02:03')
FROM ctimestamp ORDER BY a;
a IF(a < '2019-01-02 02:00:00', 'yes', 'no') ADDTIME(a, '1:1:1') STR_TO_DATE(a, '%Y-%m-%d %H:%i:%s') EXTRACT(DAY_HOUR FROM a) EXTRACT(MINUTE_SECOND FROM a) TIME_FORMAT(a, '%H:\%i:\%s') a RLIKE '02:03' IFNULL(NULL, a) CASE a WHEN '2019-01-02 01:02:03' THEN 'found' WHEN '2019-01-02 11:11:12'
THEN 'found2' ELSE 'notfound' END CHAR_LENGTH(a) CAST(a AS UNSIGNED INT) CAST(a AS CHAR) CAST(a AS DATE) TIME(CAST(a AS DATETIME)) TIME(COALESCE(NULL, a)) HEX(a) NULLIF(a, '2019-01-02 01:02:03') TIMEDIFF(a, '2019-01-01 01:02:03')
2019-01-02 01:02:03 yes 2019-01-02 02:03:04 2019-01-02 01:02:03 201 203 01:\02:\03 1 2019-01-02 01:02:03 found 19 20190102010203 2019-01-02 01:02:03 2019-01-02 01:02:03 01:02:03 323031392D30312D30322030313A30323A3033 NULL 24:00:00
2019-01-02 02:02:03 no 2019-01-02 03:03:04 2019-01-02 02:02:03 202 203 02:\02:\03 1 2019-01-02 02:02:03 notfound 19 20190102020203 2019-01-02 02:02:03 2019-01-02 02:02:03 02:02:03 323031392D30312D30322030323A30323A3033 2019-01-02 02:02:03 25:00:00
2019-01-02 11:11:12 no 2019-01-02 12:12:13 2019-01-02 11:11:12 211 1112 11:\11:\12 0 2019-01-02 11:11:12 found2 19 20190102111112 2019-01-02 11:11:12 2019-01-02 11:11:12 11:11:12 323031392D30312D30322031313A31313A3132 2019-01-02 11:11:12 34:09:09
INSERT INTO ctimestamp VALUES ('2020-01-03 12:12:12'),
('2020-05-06 12:12:12'), ('2020-10-28 12:12:12');
SELECT a, DAYNAME(a), DAYOFWEEK(a), DATE_FORMAT(a, '%W %M %Y'),
MONTHNAME(a), DATE(a), YEARWEEK(a), DAYOFYEAR(a), YEAR(a),
a + INTERVAL 1 DAY, TIMESTAMPDIFF(DAY, a, '2020-01-01'),
LAST_DAY(a), TRUNCATE(a, -2), a IN ('2019-01-02 01:02:03', a),
TO_DAYS(a), DAY(a), WEEK(a), WEEKDAY(a), GREATEST(a, '2020-07-01'),
MONTH(a), QUARTER(a), DATE_ADD(a, INTERVAL 1 SECOND)
FROM ctimestamp WHERE a > '2020-01-01' ORDER BY a;
a DAYNAME(a) DAYOFWEEK(a) DATE_FORMAT(a, '%W %M %Y') MONTHNAME(a) DATE(a) YEARWEEK(a) DAYOFYEAR(a) YEAR(a) a + INTERVAL 1 DAY TIMESTAMPDIFF(DAY, a, '2020-01-01') LAST_DAY(a) TRUNCATE(a, -2) a IN ('2019-01-02 01:02:03', a) TO_DAYS(a) DAY(a) WEEK(a) WEEKDAY(a) GREATEST(a, '2020-07-01') MONTH(a) QUARTER(a) DATE_ADD(a, INTERVAL 1 SECOND)
2020-01-03 12:12:12 Friday 6 Friday January 2020 January 2020-01-03 201952 3 2020 2020-01-04 12:12:12 -2 2020-01-31 2020-01-03 12:12:12 1 737792 3 0 4 2020-07-01 00:00:00 1 1 2020-01-03 12:12:13
2020-05-06 12:12:12 Wednesday 4 Wednesday May 2020 May 2020-05-06 202018 127 2020 2020-05-07 12:12:12 -126 2020-05-31 2020-05-06 12:12:12 1 737916 6 18 2 2020-07-01 00:00:00 5 2 2020-05-06 12:12:13
2020-10-28 12:12:12 Wednesday 4 Wednesday October 2020 October 2020-10-28 202043 302 2020 2020-10-29 12:12:12 -301 2020-10-31 2020-10-28 12:12:12 1 738091 28 43 2 2020-10-28 12:12:12 10 4 2020-10-28 12:12:13
SELECT UNIX_TIMESTAMP(a), TIME_TO_SEC(a), CEIL(a),
CAST(LEAST(a, '2019-03-03 00:00:00') AS DATETIME),
ROUND(a), SECOND(a), MINUTE(a), HOUR(a), FLOOR(a)
FROM ctimestamp ORDER BY a;
UNIX_TIMESTAMP(a) TIME_TO_SEC(a) CEIL(a) CAST(LEAST(a, '2019-03-03 00:00:00') AS DATETIME) ROUND(a) SECOND(a) MINUTE(a) HOUR(a) FLOOR(a)
1546387323 3723 2019-01-02 01:02:03 2019-01-02 01:02:03 2019-01-02 01:02:03 3 2 1 2019-01-02 01:02:03
1546390923 7323 2019-01-02 02:02:03 2019-01-02 02:02:03 2019-01-02 02:02:03 3 2 2 2019-01-02 02:02:03
1546423872 40272 2019-01-02 11:11:12 2019-01-02 11:11:12 2019-01-02 11:11:12 12 11 11 2019-01-02 11:11:12
1578049932 43932 2020-01-03 12:12:12 2019-03-03 00:00:00 2020-01-03 12:12:12 12 12 12 2020-01-03 12:12:12
1588763532 43932 2020-05-06 12:12:12 2019-03-03 00:00:00 2020-05-06 12:12:12 12 12 12 2020-05-06 12:12:12
1603883532 43932 2020-10-28 12:12:12 2019-03-03 00:00:00 2020-10-28 12:12:12 12 12 12 2020-10-28 12:12:12
DROP DATABASE IF EXISTS timestamp_test;
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a timestamp, b int);
SET time_zone='+0:00';
INSERT INTO ctimestamp VALUES ('2019-01-03 12:12:12', 1),
('2019-01-04 12:12:12', 2), ('2019-01-03 12:12:12', 4),
('2019-01-03 12:12:12', 2), ('2019-01-04 12:12:12', 1);
SELECT a, b, SUM(b) over (PARTITION BY a ORDER BY a, b
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) c
FROM ctimestamp;
a b c
2019-01-03 12:12:12 1 1
2019-01-03 12:12:12 2 3
2019-01-03 12:12:12 4 7
2019-01-04 12:12:12 1 1
2019-01-04 12:12:12 2 3
SELECT a, b, MAX(a) over (PARTITION BY b ORDER BY a desc
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) c
FROM ctimestamp;
a b c
2019-01-04 12:12:12 1 2019-01-04 12:12:12
2019-01-03 12:12:12 1 2019-01-04 12:12:12
2019-01-04 12:12:12 2 2019-01-04 12:12:12
2019-01-03 12:12:12 2 2019-01-04 12:12:12
2019-01-03 12:12:12 4 2019-01-03 12:12:12
DROP DATABASE IF EXISTS timestamp_test;
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a int, b timestamp);
INSERT INTO ctimestamp VALUES (1, 20190101), (1, 20200202),
(2, 20190202), (2, 20200202), (2, 20190101);
SELECT b, count(*) FROM ctimestamp GROUP BY b ORDER BY b;
b count(*)
2019-01-01 00:00:00 2
2019-02-02 00:00:00 1
2020-02-02 00:00:00 2
SELECT b, max(a), min(a) FROM ctimestamp GROUP BY b ORDER BY b;
b max(a) min(a)
2019-01-01 00:00:00 2 1
2019-02-02 00:00:00 2 2
2020-02-02 00:00:00 2 1
SELECT a, max(b), min(b) FROM ctimestamp GROUP BY a ORDER BY a;
a max(b) min(b)
1 2020-02-02 00:00:00 2019-01-01 00:00:00
2 2020-02-02 00:00:00 2019-01-01 00:00:00
DROP DATABASE timestamp_test;

View File

@ -0,0 +1,131 @@
--source ../include/have_columnstore.inc
--source ../include/combinations.myisam-columnstore.inc
--echo #
--echo # Test cases for the TIMESTAMP datatype
--echo #
# Test bulk insert/literals/microseconds
--disable_warnings
DROP DATABASE IF EXISTS timestamp_test;
--enable_warnings
CREATE DATABASE timestamp_test;
USE timestamp_test;
## Test the effect of changing timezones on timestamp values
CREATE TABLE ctimestamp (a timestamp);
SET time_zone='-5:00';
INSERT INTO ctimestamp VALUES ('2019-01-01 01:02:03'), ('2019-05-05 01:01:01');
SET time_zone='+1:00';
SELECT a FROM ctimestamp ORDER BY a;
SET time_zone='-2:00';
SELECT a FROM ctimestamp ORDER BY a;
## Test bulk insert using cpimport
CREATE TABLE ctimestamp2 (a timestamp DEFAULT 0);
INSERT INTO ctimestamp2 SELECT * FROM ctimestamp;
SELECT a FROM ctimestamp2 ORDER BY a;
## Test literals
CREATE TABLE ctimestamp3 (a timestamp);
INSERT INTO ctimestamp3 VALUES (19940101), (940101),
(19940101010203), (940101010203), ('1994-01-01T01:02:03');
SELECT a FROM ctimestamp3 ORDER BY a;
## Test microseconds
CREATE TABLE ctimestamp4 (a timestamp(6) default 0);
INSERT INTO ctimestamp4 VALUES (0), ('2019-01-01 01:01:01.123456');
SELECT a, microsecond(a) FROM ctimestamp4 ORDER BY a;
# Test distributed functions
--disable_warnings
DROP DATABASE IF EXISTS timestamp_test;
--enable_warnings
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a timestamp);
SET time_zone='+0:00';
INSERT INTO ctimestamp VALUES ('2019-01-02 00:02:03'),
('2019-01-02 01:02:03'), ('2019-01-02 10:11:12');
SET time_zone='+1:00';
SELECT a, a BETWEEN '2019-01-02 02:00:00' AND '2019-01-02 13:00:00'
FROM ctimestamp ORDER BY a;
SELECT a, IF(a < '2019-01-02 02:00:00', 'yes', 'no'),
ADDTIME(a, '1:1:1'), STR_TO_DATE(a, '%Y-%m-%d %H:%i:%s'),
EXTRACT(DAY_HOUR FROM a), EXTRACT(MINUTE_SECOND FROM a),
TIME_FORMAT(a, '%H:\%i:\%s'), a RLIKE '02:03', IFNULL(NULL, a),
CASE a WHEN '2019-01-02 01:02:03' THEN 'found' WHEN '2019-01-02 11:11:12'
THEN 'found2' ELSE 'notfound' END, CHAR_LENGTH(a),
CAST(a AS UNSIGNED INT), CAST(a AS CHAR), CAST(a AS DATE),
TIME(CAST(a AS DATETIME)), TIME(COALESCE(NULL, a)), HEX(a),
NULLIF(a, '2019-01-02 01:02:03'), TIMEDIFF(a, '2019-01-01 01:02:03')
FROM ctimestamp ORDER BY a;
INSERT INTO ctimestamp VALUES ('2020-01-03 12:12:12'),
('2020-05-06 12:12:12'), ('2020-10-28 12:12:12');
SELECT a, DAYNAME(a), DAYOFWEEK(a), DATE_FORMAT(a, '%W %M %Y'),
MONTHNAME(a), DATE(a), YEARWEEK(a), DAYOFYEAR(a), YEAR(a),
a + INTERVAL 1 DAY, TIMESTAMPDIFF(DAY, a, '2020-01-01'),
LAST_DAY(a), TRUNCATE(a, -2), a IN ('2019-01-02 01:02:03', a),
TO_DAYS(a), DAY(a), WEEK(a), WEEKDAY(a), GREATEST(a, '2020-07-01'),
MONTH(a), QUARTER(a), DATE_ADD(a, INTERVAL 1 SECOND)
FROM ctimestamp WHERE a > '2020-01-01' ORDER BY a;
SELECT UNIX_TIMESTAMP(a), TIME_TO_SEC(a), CEIL(a),
CAST(LEAST(a, '2019-03-03 00:00:00') AS DATETIME),
ROUND(a), SECOND(a), MINUTE(a), HOUR(a), FLOOR(a)
FROM ctimestamp ORDER BY a;
# Test window functions
--disable_warnings
DROP DATABASE IF EXISTS timestamp_test;
--enable_warnings
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a timestamp, b int);
SET time_zone='+0:00';
INSERT INTO ctimestamp VALUES ('2019-01-03 12:12:12', 1),
('2019-01-04 12:12:12', 2), ('2019-01-03 12:12:12', 4),
('2019-01-03 12:12:12', 2), ('2019-01-04 12:12:12', 1);
## Test SUM
SELECT a, b, SUM(b) over (PARTITION BY a ORDER BY a, b
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) c
FROM ctimestamp;
## Test MAX
SELECT a, b, MAX(a) over (PARTITION BY b ORDER BY a desc
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) c
FROM ctimestamp;
# Test aggregate functions
--disable_warnings
DROP DATABASE IF EXISTS timestamp_test;
--enable_warnings
CREATE DATABASE timestamp_test;
USE timestamp_test;
CREATE TABLE ctimestamp (a int, b timestamp);
INSERT INTO ctimestamp VALUES (1, 20190101), (1, 20200202),
(2, 20190202), (2, 20200202), (2, 20190101);
# Test count(*)
SELECT b, count(*) FROM ctimestamp GROUP BY b ORDER BY b;
# Test max/min
SELECT b, max(a), min(a) FROM ctimestamp GROUP BY b ORDER BY b;
SELECT a, max(b), min(b) FROM ctimestamp GROUP BY a ORDER BY a;
# Cleanup
DROP DATABASE timestamp_test;