Tzahi Fadida wrote:
Note that postgreSQL has a local swap of itself for doing big sorts etc...
However, I have a feeling its more to do with how much time it takes for
your transaction to finish. There is an automatic mechanism in postgreSQL
that drops the connection if it thinks the transaction got stuck.
I usually thinks that if in a specific command in a the plan tree the executor
cant
advance past a sub plan around 15min which is usually impossible(I think) unless
you write your own functions like me.
I don't think so because to my understanding the kill came outside
PostgreSQL ( kernel ? )
I also noticed the process worked more time after increasing shared_buffers.
Note that in postgreSQL functions can only return all their result
(materializing)
even if it seems like pipping it is not(it was not implemented yet), so all the
P lang, J lang,
etc... can get stuck.
No functions at all :) only plain SQL
Perhaps you can increase the time out, however, try first to drop the indices
in your tables (+unique constrains which are really indices) and build them
at the end.
Yes, we did that. or the process could take even more time.
We also used prepare to accelerate the insert process
besides that, as you already said, you need a bigger swap. (9GB worth :)
But thats the strange thing. the swap never get used.
Finally, this could also be a bug in postgreSQL though I doubt it.
If you write C functions like me (like a second postgreSQL in a function),
memory leaks are common :). So check your functions.
Regards,
tzahi.
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Michael Ben-Nes
Sent: Monday, February 13, 2006 9:54 AM
To: linux-il@linux.org.il
Subject: Re: RAM swap
Oleg Goldshmidt wrote:
Michael Ben-Nes [EMAIL PROTECTED] writes:
All the data is fed as one huge transaction that takes allot of
memory.
How much memory?
What do you mean ?
I noticed that one 1GB RAM machine the task cant be
completed ( kernel
kill the process )
What exactly happens? Is there anything interesting in the syslog,
such as
Out of Memory: Killed process ...
or anything of the kind?
Only Postgresql return something like:
FATAL: terminating connection due to administrator command
while on 3GB RAM it can.
The strange thing is that on the 1GB RAM machine the swap
is not used
at all.
How much swap is there?
On the 1GB there is 512MB of swap and on the 3GB there is 6GB swap
=
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word unsubscribe in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
=
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word unsubscribe in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
=
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word unsubscribe in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
