I just discovered that H2 doesn't support concurrent connections using different transaction isolation levels. Meaning, changing the transaction isolation of one connection affects all other connections.
Does Postgresql support the use of different isolation levels for each connection?
-
2Yes that works just fine. You can also cahnge the isolation level of one transaction without problems.user1822– user18222014年09月14日 06:43:43 +00:00Commented Sep 14, 2014 at 6:43
-
@a_horse_with_no_name: Thank you. I wish I could "accept" your comment.Gili– Gili2014年09月14日 06:50:31 +00:00Commented Sep 14, 2014 at 6:50
-
I have no evidence to support that, that's why I added it as a comment. As far as I can tell, there is nothing in the docs postgresql.org/docs/current/static/transaction-iso.html that would suggest such a broken behaviouruser1822– user18222014年09月14日 06:55:29 +00:00Commented Sep 14, 2014 at 6:55
-
@a_horse_with_no_name, fair enough. Thanks again for the clarification.Gili– Gili2014年09月14日 07:17:10 +00:00Commented Sep 14, 2014 at 7:17
2 Answers 2
Yes it does support different transaction isolation levels per-connection. You can set the transaction isolation level (as well as the read-only and deferrable status for transactions) for a connection with SET SESSION CHARACTERISTICS
:
localhost:5432 postgres postgres # SHOW transaction_isolation;
transaction_isolation
-----------------------
read committed
(1 row)
localhost:5432 postgres postgres # SET SESSION CHARACTERISTICS AS TRANSACTION ISOLATION LEVEL REPEATABLE READ;
SET
localhost:5432 postgres postgres # SHOW transaction_isolation;
transaction_isolation
-----------------------
repeatable read
(1 row)
You can override per-transaction with SET TRANSACTION ISOLATION LEVEL
, or as you start a transaction with BEGIN TRANSACTION ISOLATION LEVEL
:
localhost:5432 postgres postgres # BEGIN;
BEGIN
Time: 0.227 ms
localhost:5432 postgres postgres * # SET TRANSACTION ISOLATION LEVEL SERIALIZABLE;
SET
Time: 0.229 ms
localhost:5432 postgres postgres * # SHOW transaction_isolation;
transaction_isolation
-----------------------
serializable
(1 row)
Time: 0.262 ms
localhost:5432 postgres postgres * # COMMIT;
COMMIT
localhost:5432 postgres postgres # SHOW transaction_isolation;
transaction_isolation
-----------------------
repeatable read
(1 row)
The default is set with the default_transaction_isolation
parameter:
localhost:5432 postgres postgres # SHOW default_transaction_isolation;
default_transaction_isolation
-------------------------------
repeatable read
(1 row)
See the docs at http://www.postgresql.org/docs/current/static/sql-set-transaction.html
Transaction isolation level is set up per transaction basis. Even one connection can have different isolation level on each transaction.
Of course different connection can have different isolation level.
Here's some testing :
postgres=# begin;
BEGIN
postgres=# show transaction_isolation ;
transaction_isolation
-----------------------
read committed
(1 row)
postgres=# set transaction isolation level serializable ;
SET
postgres=# show transaction_isolation ;
transaction_isolation
-----------------------
serializable
(1 row)
At the same time on another session :
postgres=# begin;
BEGIN
postgres=# show transaction_isolation
;
transaction_isolation
-----------------------
read committed
(1 row)
postgres=# set transaction isolation level read uncommitted ;
SET
postgres=# show transaction_isolation
;
transaction_isolation
-----------------------
read uncommitted
(1 row)
Once You commit a transaction and begin another new transaction, default isolation level will be used which set on postgresql.conf :
#default_transaction_isolation = 'read committed'
I try this on Postgres 9.1.
Hope this help, cheers.
-
2Note that
READ UNCOMMITTED
has the same effect asREAD COMMITTED
in PostgreSQL; see postgresql.org/docs/current/static/sql-set-transaction.htmlhbn– hbn2014年09月14日 19:05:17 +00:00Commented Sep 14, 2014 at 19:05