s51512 (tools.wikihistory) is the top CPU user of toolsdb:
root@labsdb1005[information_schema]> SELECT * FROM USER_STATISTICS ORDER BY cpu_time DESC LIMIT 1\G *************************** 1. row *************************** USER: s51512 TOTAL_CONNECTIONS: 2634803 CONCURRENT_CONNECTIONS: 0 CONNECTED_TIME: 70378868 BUSY_TIME: 27301657.775641054 CPU_TIME: 22565123.800680105 BYTES_RECEIVED: 9666784385 BYTES_SENT: 19169246173 BINLOG_BYTES_WRITTEN: 3235170216 ROWS_READ: 18429748528 ROWS_SENT: 33466060 ROWS_DELETED: 6515 ROWS_INSERTED: 10179413 ROWS_UPDATED: 8192371 SELECT_COMMANDS: 51657333 UPDATE_COMMANDS: 17789158 OTHER_COMMANDS: 2634765 COMMIT_TRANSACTIONS: 77495286 ROLLBACK_TRANSACTIONS: 4857285 DENIED_CONNECTIONS: 0 LOST_CONNECTIONS: 0 ACCESS_DENIED: 0 EMPTY_QUERIES: 19721750 1 row in set (0.00 sec)
taking half of the total resources, in a database that serves hundreds to thousands of accounts:
root@labsdb1005[information_schema]> SELECT sum(cpu_time) FROM USER_STATISTICS; +-------------------+ | sum(cpu_time) | +-------------------+ | 40725292.63999812 | +-------------------+ 1 row in set (0.00 sec) root@labsdb1005[information_schema]>
This is specially important because as far as I can tell, there is 8 threads doing highly inneficient and easily optimazable queries (Using filesort on a lot of rows):
root@labsdb1005[(none)]> SHOW EXPLAIN FOR 63790240\G *************************** 1. row *************************** id: 1 select_type: SIMPLE table: dewiki_data type: range possible_keys: queue_sorting key: queue_sorting key_len: 9 ref: NULL rows: 2451785 Extra: Using index condition; Using filesort 1 row in set, 1 warning (0.26 sec)
Because of this, CPU is at 100% percent, leaving no resources for other users:
Please make the queries more efficient, reduce concurrency or ask for dedicated resources to avoid affecting other users. I ended up doing an analysis like this because a user complained of slowdown on toolsdb on IRC.