WAL Files checkpoint_timeout with voluminous delete/insert

[Hervé Piedvache]

Hi,

A small question ... and your opinions are well done ...

I have tested one thing today ...
I have a script in charge to insert a lot of files inside one of my db.
I have a perl script where I read the files ... check if some similar data are
in the db ... then delete them ... and insert the contains of the file.
I do this file by file, and each time I make a transaction where I delete the
data first (one delete command) and insert the new data with one insert by
line ... then I commit and check the next file ...
My file have between 200 to 5000 lines each times ... in my test I have about
650 000 lignes in total ...
I have a dedicated partition on the disk for the WAL files, 2Gb.

So I start my script and saw the size of the pg_xlog partition groing quickly
to be full after 20 min ...

So ... I think it's my checkpoint_timeout who is too large ... I look at my
config file and I see :
#checkpoint_timeout = 300

So default ... it's not set ... and the default value seems to be 300 ... (I
have tested after).

My question is simple ... if I reduce the time of the checkpoint_timeout ...
what is the impact in normal use of my database ... does someone have done
any test ? ...

Is it possible in futur version to change somewhere in realtime this value
just for big insert like this ... or this idea have no sence ?

Thanks for your reply,

Regards,
--
Hervé Piedvache

Elma Ingénierie Informatique
6 rue du Faubourg Saint-Honoré
F-75008 - Paris - France
Pho. 33-144949901
Fax. [phone]
---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match

[Tom Lane]

=?iso-8859-15?q?Herv=E9=20Piedvache?= <he***@elma.fr> writes:

I have a dedicated partition on the disk for the WAL files, 2Gb. So I start my script and saw the size of the pg_xlog partition groing quickly
to be full after 20 min ...

That shouldn't happen, unless there's something preventing automatic
checkpoints from occurring. What do you have checkpoint_segments set
to?

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings

[Hervé Piedvache]

Hi Tom,

Le Vendredi 22 Août 2003 22:43, Tom Lane a écrit :

=?iso-8859-15?q?Herv=E9=20Piedvache?= <he***@elma.fr> writes:

I have a dedicated partition on the disk for the WAL files, 2Gb.

So I start my script and saw the size of the pg_xlog partition groing
quickly to be full after 20 min ...

That shouldn't happen, unless there's something preventing automatic
checkpoints from occurring. What do you have checkpoint_segments set
to?

#checkpoint_segments = 3 # in logfile segments, min 1, 16MB each

Unset in the config file ... so default value ... 3 ?

So, any idea of my problem ?

regards,
--
Hervé Piedvache

Elma Ingénierie Informatique
6 rue du Faubourg Saint-Honoré
F-75008 - Paris - France
Pho. 33-144949901
Fax. [phone]
---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
(send "unregister YourEmailAddressHere" to ma*******@postgresql.org)