CODEWITHSUNDEEP
mysqlinnodbdatabase-deadlocks
ThiefMaster
ThiefMaster

Reputation: 318468

Deadlock found in MySQL (InnoDB)

I'm getting the following error pretty often when running a certain query in my database (all tables use the InnoDB storage engine): "Deadlock found when trying to get lock; try restarting transaction"

The query is DELETE FROM sessions WHERE userid != 0 AND lastactivity < 1289594761 AND admin = 1 AND userid NOT IN (SELECT userid FROM users WHERE (userflags & 1048576))

The errors started to occur when I've added the NOT IN part to my WHERE statement. Why is this causing issues and what can I do to prevent this?

Upvotes: 0

Views: 882

Answers (2)

bobince
bobince

Reputation: 536329

Presumably you get the error more often because this is now a much slower query.

The & op on userflags makes the subquery unindexable. Flag words aren't usually good schema design, as they require computation that defeats indexing. If you are doing bit testing queries a lot, separate columns of small datatypes (eg TINYINT) may be better.

If your schema works the way it looks like it might, you should be able to do this using a simple JOIN, which usually performs better than a subqueries:

DELETE sessions
FROM sessions
JOIN users ON users.userid=sessions.userid
WHERE sessions.lastactivity<1289594761 AND admin=1
AND (users.userflags&1048576)=0

(joins on DELETE is a non-ANSI SQL extension in MySQL.)

Upvotes: 2

Cahit
Cahit

Reputation: 2534

An easy solution would be to separate this into two consecutive queries. I.e.,:

SELECT userid into #tmptable FROM users WHERE (userflags & 1048576);

DELETE FROM sessions WHERE userid != 0 AND lastactivity < 1289594761 AND admin = 1 AND userid NOT IN (select userid from #tmptable);

That way you're working with a local session copy of the values from the second table and not causing a read lock on it. However, this is just a quick-and-dirty solution. A better solution would be to analyze the transaction lock settings from all activities that touch these two tables and to rewrite the query, if you'll be re-using it regularly.

Upvotes: 2

Related Questions