Portál AbcLinuxu, 7. května 2025 06:06
Zdravim, mam MySQL v 3 uzlech pomoci master-master replikace Percona xtradb cluster.
Vse beha v poho az do chvile kdy po nejake dobe insertne zaznam do db ktera ma neco kolem 300 zaznamu. V tuto chvili se postupne zatizi vsechny tri uzle replikace a prestanou reagovat na par desitek vterin repsektive kousne se cela aplikace.
Ale nad touto tabulkou se provadi relativne dost SELECTu radove desitky za vterinu.
Myslim ze to muze delat velka hodnota query-cache-size ale s mensi hodnotou bylo zase hodne prunes per day.
Dokaze nekdo prosim nakopnout spravnym smerem co hledat?
V logu mam pak neco v tomto duchu (Uvedena table v logu nize nesouvisi prave s tou ve ktere se provedl insert)
*** Priority TRANSACTION: TRANSACTION 393262833, ACTIVE 0 sec starting index read mysql tables in use 1, locked 1 1 lock struct(s), heap size 360, 0 row lock(s) MySQL thread id 7, OS thread handle 0x7f255962e700, query id 299104427 invalidating query cache entries (table) *** Victim TRANSACTION: TRANSACTION 393257917, ACTIVE 114 sec mysql tables in use 1, locked 1 2 lock struct(s), heap size 360, 1 row lock(s), undo log entries 1 MySQL thread id 25482552, OS thread handle 0x7f252e134700, query id 299095466 localhost 127.0.0.1 database_sys wsrep in pre-commit stage UPDATE `table` SET `record`=1.5655388889 WHERE (`table`.`id` = 10776) *** WAITING FOR THIS LOCK TO BE GRANTED: RECORD LOCKS space id 4039 page no 126 n bits 176 index `PRIMARY` of table `database`.`table` trx id 393257917 lock_mode X locks rec but not gap 2015-02-26 22:47:53 58270 [Note] WSREP: cluster conflict due to high priority abort for threads: 2015-02-26 22:47:53 58270 [Note] WSREP: Winning thread: THD: 7, mode: applier, state: executing, conflict: no conflict, seqno: 33153378 SQL: (null) 2015-02-26 22:47:53 58270 [Note] WSREP: Victim thread: THD: 25482552, mode: local, state: committing, conflict: no conflict, seqno: -1 SQL: UPDATE `table` SET `record`=1.5655388889 WHERE (`table`.`id` = 10776)
Pro MySQL je nasledujici konfigurace
[mysqld]
innodb_use_native_aio=0
datadir = /var/lib/mysql
pid-file = /var/run/mysqld/mysqld.pid
key-buffer-size = 32M
myisam-recover = FORCE,BACKUP
max-allowed-packet = 16M
max-connect-errors = 1000000
innodb = FORCE
tmp-table-size = 32M
max-heap-table-size = 32M
max-connections = 500
thread-cache-size = 50
open-files-limit = 65535
table-definition-cache = 1024
table-open-cache = 2048
log-error = /var/log/mysql/error.log
log-queries-not-using-indexes = 0
slow-query-log = 0
slow-query-log-file = /var/log/mysql/slow.log
# CACHES AND LIMITS #
#table_cache = 16000
tmp-table-size = 256M
max-heap-table-size = 256M
query-cache-type = 1
query-cache-size = 2048M
max-connections = 2000
thread-cache-size = 50
open-files-limit = 65535
table-definition-cache = 8000
table-open-cache = 8000
key_buffer_size = 256M
join_buffer_size = 512k
thread_cache_size = 15000
table_definition_cache = 16384
table_open_cache = 16384
innodb_write_io_threads = 16
innodb_read_io_threads = 16
back_log = 2048
# INNODB #
innodb_lock_wait_timeout = 60
innodb_buffer_pool_instances = 8
innodb_rollback_on_timeout = 0
innodb-flush-method = O_DIRECT
innodb-log-files-in-group = 2
innodb-log-file-size = 128M
#innodb-flush-log-at-trx-commit = 1
innodb-file-per-table = 1
innodb-buffer-pool-size = 4096M
innodb-flush-log-at-trx-commit = 0
[--] Up for: 18d 2h 26m 48s (258M q [165.038 qps], 23M conn, TX: 164B, RX: 39B)
[--] Reads / Writes: 69% / 31%
[--] Total buffers: 6.5G global + 1.4M per thread (800 max threads)
[OK] Maximum possible memory usage: 7.6G (48% of installed RAM)
[OK] Slow queries: 0% (4K/258M)
[OK] Highest usage of available connections: 34% (275/800)
[OK] Key buffer size / total MyISAM indexes: 256.0M/102.0K
[OK] Key buffer hit rate: 100.0% (171M cached / 4 reads)
[OK] Query cache efficiency: 46.5% (89M cached / 192M selects)
[OK] Query cache prunes per day: 0
[OK] Sorts requiring temporary tables: 0% (3K temp sorts / 15M sorts)
[OK] Temporary tables created on disk: 0% (1K on disk / 11M total)
[OK] Thread cache hit rate: 99% (308 created / 23M connections)
[OK] Table cache hit rate: 87% (3K open / 4K opened)
[OK] Open file limit used: 0% (18/33K)
[OK] Table locks acquired immediately: 99% (166M immediate / 166M locks)
[!!] InnoDB buffer pool / data size: 4.0G/5.0G
[OK] InnoDB log waits: 0
[--] Up for: 18d 2h 26m 48s (258M q [165.038 qps], 23M conn, TX: 164B, RX: 39B)
[--] Reads / Writes: 69% / 31%
[--] Total buffers: 6.5G global + 1.4M per thread (800 max threads)
[OK] Maximum possible memory usage: 7.6G (48% of installed RAM)
[OK] Slow queries: 0% (4K/258M)
[OK] Highest usage of available connections: 34% (275/800)
[OK] Key buffer size / total MyISAM indexes: 256.0M/102.0K
[OK] Key buffer hit rate: 100.0% (171M cached / 4 reads)
[OK] Query cache efficiency: 46.5% (89M cached / 192M selects)
[OK] Query cache prunes per day: 0
[OK] Sorts requiring temporary tables: 0% (3K temp sorts / 15M sorts)
[OK] Temporary tables created on disk: 0% (1K on disk / 11M total)
[OK] Thread cache hit rate: 99% (308 created / 23M connections)
[OK] Table cache hit rate: 87% (3K open / 4K opened)
[OK] Open file limit used: 0% (18/33K)
[OK] Table locks acquired immediately: 99% (166M immediate / 166M locks)
[!!] InnoDB buffer pool / data size: 4.0G/5.0G
[OK] InnoDB log waits: 0
17.3.2015 16:18
rADOn | skóre: 44
| blog: bloK
| Praha
Nejjednodušším způsobem, jak se vyhýbat deadlocku, je zamykání tabulek vždy ve stejném pořadí, či používání enginu InnoDB, který deadlocky sám detekuje.InnoDB mam vsude bez vyhrady, vyuzivam Foreign keys..
21.3.2015 13:12
okbob | skóre: 30
| blog: systemakuv_blog
| Benešov
mysql> SHOW GLOBAL STATUS LIKE 'innodb_deadlocks'; +------------------+-------+ | Variable_name | Value | +------------------+-------+ | Innodb_deadlocks | 0 | +------------------+-------+ 1 row in set (0,11 sec)
21.3.2015 18:44
okbob | skóre: 30
| blog: systemakuv_blog
| Benešov
Kdyz jsem na to koukal tak deadlocky se stavaji na tabulkach ktery by nemeli mit s tim nic spolecneho..
Problem nastane kdyz pridam zaznam do tabulky se seznamem uzivatelskych blokaci (ktere se nactou pri kazdem hitu prihlaseneho uzivatele), ma 4 cizi klice (2 z toho vedou na stejnou tabulku uzivatelu ktera se prakticky neaktualizuje + jeden ktery je ne-vzdy vyplneny vede na tabulku zaznamu pokud se blokace zapisuje jeste do zvlastni tabulky tzn blokace je vyssi urovne administratorem a posledni klic vede na tabulku tez taky se seznamem ktery neni aktualizovan tak casto ani a hlavne ne pri vlozeni blokace). PK je pouze na ID sloupec, tzn PK + 4 IK s FK.
Zkusim se mrknout na ten link a povolim verbose logovani locku snad z toho budu moudrejsi jsem z toho uz trochu zmateny, k percone jsem nasel i toto v toolkitu - http://www.percona.com/doc/percona-toolkit/2.1/pt-deadlock-logger.html
Po nejakem case tu mam maly update. Nyni mi percona jede single v jednom boostrap-pxc uzlu a problem s tou tabulkou je neustale. Takze timto odpada teorie o problemu v (nebo při) replikaci
Nize posilam schema tabulky. 3 IK+FK a 1 PK s AI. Aktualni pocet zaznamu 387. Kdyz to necham pul dne odlezet a udelam insert tak mysql prestane reagovat na 10s. Kdyz pote udelam hned dalsi tak vse ok.
CREATE TABLE `mistnosti_zakazy` ( `id` INT(11) NOT NULL AUTO_INCREMENT, `uzivatel_id` INT(11) NULL DEFAULT NULL, `mistnost_id` INT(11) NULL DEFAULT NULL, `ip` VARCHAR(15) NULL DEFAULT NULL COLLATE 'utf8_czech_ci', `kolikrat` TINYINT(4) NULL DEFAULT '0', `typ` INT(11) NULL DEFAULT NULL, `duvod` VARCHAR(1000) NULL DEFAULT NULL COLLATE 'utf8_czech_ci', `od_id` INT(11) NULL DEFAULT NULL, `do` DATETIME NULL DEFAULT NULL, `trest_id` INT(11) NULL DEFAULT NULL, PRIMARY KEY (`id`), INDEX `FK__uzivatele` (`uzivatel_id`), INDEX `FK__mistnosti` (`mistnost_id`), INDEX `FK_mistnosti_zakazy_uzivatele` (`od_id`), INDEX `FK_mistnosti_zakazy_uzivatele_tresty` (`trest_id`), CONSTRAINT `FK_mistnosti_zakazy_mistnosti` FOREIGN KEY (`mistnost_id`) REFERENCES `mistnosti` (`id`) ON UPDATE CASCADE ON DELETE CASCADE, CONSTRAINT `FK_mistnosti_zakazy_uzivatele` FOREIGN KEY (`uzivatel_id`) REFERENCES `uzivatele` (`id`) ON UPDATE CASCADE ON DELETE CASCADE, CONSTRAINT `FK_mistnosti_zakazy_uzivatele_2` FOREIGN KEY (`od_id`) REFERENCES `uzivatele` (`id`) ON UPDATE CASCADE ON DELETE CASCADE, CONSTRAINT `FK_mistnosti_zakazy_uzivatele_tresty` FOREIGN KEY (`trest_id`) REFERENCES `uzivatele_tresty` (`id`) ON UPDATE CASCADE ON DELETE CASCADE ) COLLATE='utf8_czech_ci' ENGINE=InnoDB AUTO_INCREMENT=11039 ;
Tiskni
Sdílej:
ISSN 1214-1267, (c) 1999-2007 Stickfish s.r.o.