Partitioned tables and locks

Started by James Sewellover 6 years ago4 messagesgeneral
Jump to latest
#1James Sewell
james.sewell@jirotech.com

Is it expected that a lock on a partitioned table will take out 2 locks per
child regardless of the number of children which are excluded at plan time?

For example I can select count(*) from a table with 3500 partitions in
transaction, and see 7000 AccessShareLocks show up till I finish the
transaction.

Bonus question - sometimes if I query pg_locks fast enough I can see ~30K
locks - how could this possibly be the case. This is on an otherwise idle
system.

Cheers,
James

--
The contents of this email are confidential and may be subject to legal or
professional privilege and copyright. No representation is made that this
email is free of viruses or other defects. If you have received this
communication in error, you may not copy or distribute any part of it or
otherwise disclose its contents to anyone. Please advise the sender of your
incorrect receipt of this correspondence.

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: James Sewell (#1)
Re: Partitioned tables and locks

James Sewell <james.sewell@jirotech.com> writes:

Is it expected that a lock on a partitioned table will take out 2 locks per
child regardless of the number of children which are excluded at plan time?

Depends on the details of your query, and on which PG version you're
using, but it's by no means surprising for each child table to get
locked. (I'm not sure where *two* locks would come from, though.)

If you're working with massively partitioned tables, increasing
max_locks_per_transaction is a good idea.

regards, tom lane

#3James Sewell
james.sewell@jirotech.com
In reply to: Tom Lane (#2)
Re: Partitioned tables and locks

Is it expected that a lock on a partitioned table will take out 2 locks
per

child regardless of the number of children which are excluded at plan

time?

Depends on the details of your query, and on which PG version you're
using, but it's by no means surprising for each child table to get
locked. (I'm not sure where *two* locks would come from, though.)

The queries were just select count(*) both with a where clause which
excludes some partitions and without, I’m on 11.

The table has over 3K partitions

If you're working with massively partitioned tables, increasing
max_locks_per_transaction is a good idea.

It’s 256 at the moment, with 600 connections - guess I need more.
--
James Sewell,
Chief Architect

Suite 112, Jones Bay Wharf, 26-32 Pirrama Road, Pyrmont NSW 2009
P (+61) 2 8099 9000 W www.jirotech.com F (+61) 2 8099 9099

--
The contents of this email are confidential and may be subject to legal or
professional privilege and copyright. No representation is made that this
email is free of viruses or other defects. If you have received this
communication in error, you may not copy or distribute any part of it or
otherwise disclose its contents to anyone. Please advise the sender of your
incorrect receipt of this correspondence.

#4Michael Lewis
mlewis@entrata.com
In reply to: James Sewell (#3)
Re: Partitioned tables and locks

I don't recall the details, but I know v12 included significant
enhancements to lock relations later in the process such that when
targeting relatively few of the partitions, it can be a major performance
boost.