Re: Deadlocks in 7.4.x ... - Mailing list pgsql-hackers

From Tom Lane
Subject Re: Deadlocks in 7.4.x ...
Date
Msg-id 16906.1116908610@sss.pgh.pa.us
Whole thread Raw
In response to Deadlocks in 7.4.x ...  ("Marc G. Fournier" <scrappy@postgresql.org>)
List pgsql-hackers
"Marc G. Fournier" <scrappy@postgresql.org> writes:
> ...
>    3) All deadlocks are fighting for the same relation: 4335271, which is
>       the tsearch2 fulltext index on the articles table:
>       articles_idxft1_idx.

> DETAIL:  Process 8122 waits for AccessExclusiveLock on relation 4335271 of database 17142; blocked by process 11846.
> Process 11846 waits for ShareLock on transaction 2897133915; blocked by process 10042.
> Process 10042 waits for AccessExclusiveLock on relation 4335271 of database 17142; blocked by process 8122.
> ----------------------------------
> DETAIL:  Process 10042 waits for AccessExclusiveLock on relation 4335271 of database 17142; blocked by process
11846.
> Process 11846 waits for ShareLock on transaction 2897133915; blocked by process 10042.

Gist indexes require AccessExclusiveLock for any update, so the blocks
on 4335271 just indicate an index update attempt.  The "ShareLock on
transaction" lines suggest a block on a row that is updated or selected
FOR UPDATE.  It's hard to say more without more info.  Do you have any
related foreign keys, for instance?
        regards, tom lane


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: Speeding up the Postgres lexer
Next
From: Markus Bertheau
Date:
Subject: Re: [SQL] ARRAY() returning NULL instead of ARRAY[] resp. {}