Last modified: 2014-10-24 08:33:16 UTC
Different results with queries in labs versus production around september 21st. The following query returns different (very different) results in production that it does in labs: Production: ---------------- SELECT count(log_user) FROM enwiki.logging /* exclude proxy registrations */ WHERE log_type = 'newusers' /* only include self-created users, exclude attached and proxy-registered users */ AND log_action = 'create' AND log_timestamp BETWEEN 20140921000000 AND 20140922000000; Returns: 8027 Labs: -------- SELECT count(log_user) FROM logging /* exclude proxy registrations */ WHERE log_type = 'newusers' /* only include self-created users, exclude attached and proxy-registered users */ AND log_action = 'create' AND log_timestamp BETWEEN 20140921000000 AND 20140922000000; Returns: 6842 Halfak did some digging and placed the missing rows in analytics-store:staging.missing_labs_new_user_20140921 It looks like there's 5 hours of the day where the rows were missing. mysql:research@analytics-store.eqiad.wmnet [staging]> select LEFT(log_timestamp, 10) as hour, count(*) from missing_labs_new_user_20140921 GROUP BY 1; +------------+----------+ | hour | count(*) | +------------+----------+ | 2014092108 | 83 | | 2014092109 | 336 | | 2014092110 | 304 | | 2014092111 | 344 | | 2014092112 | 118 | +------------+----------+ 5 rows in set (0.01 sec)
Sync in progress. Cause is not yet confirmed, with https://mariadb.atlassian.net/browse/MDEV-6551 a possibility. Very interested to hear if anyone observes this with recent (<1week) data.