pg.ddx.io  pgsql-general@postgresql.org mailing list archive  
help / color / mirror / Atom feed
Question about tuple´s lock
4+ messages / 2 participants
[nested] [flat]

* Question about tuple´s lock
@ 2023-10-24 16:27  Carlos Alves <carlos.alves@tecnisys.com.br>
  0 siblings, 2 replies; 4+ messages in thread

From: Carlos Alves @ 2023-10-24 16:27 UTC (permalink / raw)
  To: pgsql-general@lists.postgresql.org

Hi! 

I need some help to understand why a transaction wiht a row delete
aquire a AccessExclusiveLock and a row update a ExclusiveLock. 

To ilustrate, a made this scenario: 

create table tblock 

( 

  id int primary key, 

  value text 

); 

insert into tblock values (1, 'somevalue'); 

First start Transaction 1: 

begin; 

update tblock set value = 'othervalue' where id = 1; 

-- keep this transaction opened 

After, start Transaction 2: 

begin; 

update tblock set value = 'onemorevalue' where id = 1; 

-- at this point, this transaction assumes a ExclusiveLock in the tuple;


-- keep this transaction opend 

Then, start Transaction 3: 

begin; 

update tblock set value = 'lastofthevalues' where id = 1; 

-- here i can see this transaction trying to aquire a
AccessExclusiveLock in this tuple 

So, my question is: why a delete row requires a AccessExclusiveLock in
the tuple insteead of ExclusiveLock? 

Bellow, a image of a query over pg_locks: 

PostgreSQL version: 13 

OS: Centos 7 

Thanks in advance.

-- 

 Carlos Alves 
Especialista

Sia Trecho 08, lotes 245 / 255 / 265 || +55 (61) 3039-9700
71205-080 || Guará || Brasília, DF 0800-6020097 

www.tecnisys.com.br [1] 

 

Links:
------
[1] http://www.tecnisys.com.br

Attachments:

b9879537.png (image/png, 73.1K) download view image
f62593aa.gif (image/gif, 43B) download view image
04237a95.gif (image/gif, 3.1K) download view image
^ permalink raw reply [nested|flat] 4+ messages in thread

* Re: Question about tuple´s lock
@ 2023-10-24 16:33  Carlos Alves <carlos.alves@tecnisys.com.br>
  parent: Carlos Alves <carlos.alves@tecnisys.com.br>
  1 sibling, 0 replies; 4+ messages in thread

From: Carlos Alves @ 2023-10-24 16:33 UTC (permalink / raw)
  To: pgsql-general@lists.postgresql.org

Em 24/10/2023 13:27, Carlos Alves escreveu:

> Hi! 
> 
> I need some help to understand why a transaction wiht a row delete aquire a AccessExclusiveLock and a row update a ExclusiveLock. 
> 
> To ilustrate, a made this scenario: 
> 
> create table tblock 
> 
> ( 
> 
> id int primary key, 
> 
> value text 
> 
> ); 
> 
> insert into tblock values (1, 'somevalue'); 
> 
> First start Transaction 1: 
> 
> begin; 
> 
> update tblock set value = 'othervalue' where id = 1; 
> 
> -- keep this transaction opened 
> 
> After, start Transaction 2: 
> 
> begin; 
> 
> update tblock set value = 'onemorevalue' where id = 1; 
> 
> -- at this point, this transaction assumes a ExclusiveLock in the tuple; 
> 
> -- keep this transaction opend 
> 
> Then, start Transaction 3: 
> 
> begin; 
> 
> update tblock set value = 'lastofthevalues' where id = 1; 
> 
> -- here i can see this transaction trying to aquire a AccessExclusiveLock in this tuple 
> 
> So, my question is: why a delete row requires a AccessExclusiveLock in the tuple insteead of ExclusiveLock? 
> 
> Bellow, a image of a query over pg_locks: 
> 
> PostgreSQL version: 13 
> 
> OS: Centos 7 
> 
> Thanks in advance.

I typed the last command wrong. Should be:

Transaction 3: 

begin; 

delete from tblock where id = 1; 

Sorry!

Attachments:

b9879537.png (image/png, 73.1K) download view image
f62593aa.gif (image/gif, 43B) download view image
04237a95.gif (image/gif, 3.1K) download view image
^ permalink raw reply [nested|flat] 4+ messages in thread

* Re: Question about tuple´s lock
@ 2023-10-24 20:31  Tom Lane <tgl@sss.pgh.pa.us>
  parent: Carlos Alves <carlos.alves@tecnisys.com.br>
  1 sibling, 1 reply; 4+ messages in thread

From: Tom Lane @ 2023-10-24 20:31 UTC (permalink / raw)
  To: Carlos Alves <carlos.alves@tecnisys.com.br>; +Cc: pgsql-general@lists.postgresql.org

Carlos Alves <carlos.alves@tecnisys.com.br> writes:
> I need some help to understand why a transaction wiht a row delete
> aquire a AccessExclusiveLock and a row update a ExclusiveLock. 

UPDATE can use the weaker lock type if it's not modifying any
column that is part of a unique index.  This is to allow concurrency
with foreign-key checks that might wish to grab a read-only (shared)
lock on such a tuple.

A DELETE, or an UPDATE that is modifying key columns, has to
conflict with foreign-key checks.

			regards, tom lane





^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: Question about tuple´s lock
@ 2023-10-24 20:43  Carlos Alves <carlos.alves@tecnisys.com.br>
  parent: Tom Lane <tgl@sss.pgh.pa.us>
  0 siblings, 0 replies; 4+ messages in thread

From: Carlos Alves @ 2023-10-24 20:43 UTC (permalink / raw)
  To: Tom Lane <tgl@sss.pgh.pa.us>; +Cc: pgsql-general@lists.postgresql.org

Em 24/10/2023 17:31, Tom Lane escreveu:

> Carlos Alves <carlos.alves@tecnisys.com.br> writes: 
> 
>> I need some help to understand why a transaction wiht a row delete
>> aquire a AccessExclusiveLock and a row update a ExclusiveLock.
> 
> UPDATE can use the weaker lock type if it's not modifying any
> column that is part of a unique index.  This is to allow concurrency
> with foreign-key checks that might wish to grab a read-only (shared)
> lock on such a tuple.
> 
> A DELETE, or an UPDATE that is modifying key columns, has to
> conflict with foreign-key checks.
> 
> regards, tom lane

Tom, 

thank you very much for your straight to the point answer!

regards
Carlos Alves

^ permalink  raw  reply  [nested|flat] 4+ messages in thread


end of thread, other threads:[~2023-10-24 20:43 UTC | newest]

Thread overview: 4+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2023-10-24 16:27 Question about tuple´s lock Carlos Alves <carlos.alves@tecnisys.com.br>
2023-10-24 16:33 ` Carlos Alves <carlos.alves@tecnisys.com.br>
2023-10-24 20:31 ` Tom Lane <tgl@sss.pgh.pa.us>
2023-10-24 20:43   ` Carlos Alves <carlos.alves@tecnisys.com.br>

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox