Re: transactions getting slon in councurrent environment - Mailing list pgsql-bugs

From Russell Smith
Subject Re: transactions getting slon in councurrent environment
Date
Msg-id 45958D99.5040506@pws.com.au
Whole thread Raw
In response to Re: transactions getting slon in councurrent environment  (Tigran Mkrtchyan <tigran.mkrtchyan@desy.de>)
List pgsql-bugs
Tigran Mkrtchyan wrote:
> Does it mean that I have to commit after each select statement?
>
> Here what the manual says:
>
> ------------
> Description
>
> COMMIT commits the current transaction. All changes made by the
> transaction become visible to others and are guaranteed to be durable
> if a crash occurs.
> -----------
>
> Does select produces some changes?
No, but with autocommit off, you are effectively doing
START TRANSACTION;
SELECT * from x;

and then leaving the transaction open.  The way MVCC works is to write
new data into the heap when you insert/update.  It doesn't replace the
data.  As this select transaction is open, it may still need to read
data from before the update.  So VACUUM can't clean that data out of the
table when it runs and the data still needs to be available to that
transaction.

So basically if you have autocommit off, the transaction stays open.
You can either set autocommit on, or you can COMMIT the transaction.

Russell.
>
>
> Regards,
>     Tigran.
>
> Tom Lane wrote:
>> Tigran Mkrtchyan <tigran.mkrtchyan@desy.de> writes:
>>> I got a incorrect or unexpected behavior in concurrent environment.
>>
>> This is not a bug, nor even surprising.  Since you haven't committed
>> the second transaction, there are a growing number of
>> dead-but-not-recyclable versions of the updated row.  The active client
>> has to check each of these versions during its primary key uniqueness
>> check during each update.
>>
>>             regards, tom lane
>
>

pgsql-bugs by date:

Previous
From: Tigran Mkrtchyan
Date:
Subject: Re: transactions getting slon in councurrent environment
Next
From: "George"
Date:
Subject: BUG #2868: Help