Page MenuHomePhabricator

Many INSERT queries on some web requests
Closed, DuplicatePublic

Description

Still seeing things like this occasionally:

2014-07-24 18:01:35 mw1191 outreachwiki: Sub-optimal transaction on DB(s) [10.64.16.27 (outreachwiki) (TRX#9ebd0c667c2b)]:
0 0.003455 query-m: UPDATE page SET page_links_updated = 'X' [TRX#9ebd0c667c2b]
1 0.000335 query-m: BEGIN
2 0.002425 query-m: DELETE FROM protected_titles WHERE pt_namespace = 'X' [TRX#9ebd0c667c2b]
3 0.000422 query-m: REPLACE INTO updates (up_page,up_action,up_timestamp,up_sequence) VALUES ('X',NULL) [TRX#9ebd0c667c2b]
4 0.002569 query-m: DELETE FROM revtag WHERE rt_page = 'X' [TRX#9ebd0c667c2b]
5 0.001807 query-m: DELETE FROM translate_groupstats WHERE tgs_group = 'X' [TRX#9ebd0c667c2b]
6 0.001250 query-m: REPLACE INTO revtag (rt_page,rt_type,rt_revision,rt_value) VALUES ('X') [TRX#9ebd0c667c2b]
7 0.000414 query-m: INSERT IGNORE INTO translate_groupstats (tgs_group,tgs_lang,tgs_total,tgs_translated,tgs_fuzzy,tgs_proofread) VALUES ('X') [TRX#9ebd0c667c2b]
8 0.000831 query-m: INSERT IGNORE INTO translate_groupstats (tgs_group,tgs_lang,tgs_total,tgs_translated,tgs_fuzzy,tgs_proofread) VALUES ('X') [TRX#9ebd0c667c2b]
9 0.004765 query-m: INSERT IGNORE INTO translate_groupstats (tgs_group,tgs_lang,tgs_total,tgs_translated,tgs_fuzzy,tgs_proofread) VALUES ('X') [TRX#9ebd0c667c2b]
10 0.005821 query-m: INSERT IGNORE INTO translate_groupstats (tgs_group,tgs_lang,tgs_total,tgs_translated,tgs_fuzzy,tgs_proofread) VALUES ('X') [TRX#9ebd0c667c2b]
...
366 0.005821 query-m: INSERT IGNORE INTO translate_groupstats (tgs_group,tgs_lang,tgs_total,tgs_translated,tgs_fuzzy,tgs_proofread) VALUES ('X') [TRX#9ebd0c667c2b]

There must have been some >= 3 second method but the log entry was cutoff (being too long). Can these sorts of large updates always use the job queue?

Details

Reference
bz68530

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 22 2014, 3:31 AM
bzimport set Reference to bz68530.
bzimport added a subscriber: Unknown Object (MLST).
aaron created this task.Jul 24 2014, 8:04 PM

I am wondering what should be the priority of this task now. It is still not using JobQueue, but the writes are batched into one idle query.