We are currently trying to use PostgreSQL in a high throuput OLTP style
Without trying to give you all the details of what we are doing(There is
quite a bit) I just want to explain the problem and in general what we are
doing. I am sure this will be something you have come accrross before.
We install PostgreSQL, install our schema and stored procedures.
We then populate the db with lots of test data.
Perform a vacuumdb db to generate stats
Remove all the test data.
Stop the server.
We then tar up all the files.
Fine so far, this is our set of files for deployment.
We deploy the db and the rest of our applications and start them all up.
Our application then starts and send lots of transactions to the db over
three separate transactions.
It takes between 200 and 2000 ms for PostgreSQL to service each of our
It does not seem to matter how long we leave the thing running or how many
time we vacuum thi is how long transactions take.
We then stop the db and our application,
Start the db by itself
remove all data
start our apllication
The the exact same transactions that took 200 to 2000 ms now take between 20
and 40 ms.
This happens on nearly every deployment but I can not find or understand the
root of the problem.
Thanks in advance.
pgsql-bugs by date
|Next:||From: Scott Shattuck||Date: 2002-07-23 15:00:17|
|Subject: Re: VACUUM FULL changes the order of rows in a table?|
|Previous:||From: eutm||Date: 2002-07-23 08:21:41|
|Subject: Patch for Re: [HACKERS] Bug of PL/pgSQL parser|