I think I've confused the issue a bit and the aforementioned time hit might be normal given: http://www.sqlite.org/lockingv3.html
We are using a separately developed COM DLL interface to Sqlite v3: http://www.sqliteplus.com/ And we can live with the journal creation time if that is what it is. Tom ---------------------------------------- From: [EMAIL PROTECTED] Sent: Monday, December 19, 2005 6:43 AM To: sqlite-users@sqlite.org Subject: [sqlite] multiple Db's and journal file time hit? Hello Recently it was written: whenever a modifying transaction starts (or rather, when it has to actually write data to disk for the first time), SQLite has to create a journal file for each DB involved as well as a master journal file. We are testing (almost ready to purchase the $180 COM version) and have just noticed a 30-40 second hit at early on in our program running. It happens when accessing the databases but not always the same call and we are wondering if it might be related to journal file start up time OR if there is some kind of built-in pause or delay in the demo version. Background: a) We use two databases- one for a current run and one for long-term history. b) The historical database has two tables with two indexes (an integer (hash code) and a date column) and is set to synchronous = Normal c) The current-run database has one table with two indexes (an integer (hash code) and another integer) and is set to synchronous = OFF. Note: we blow away the current-run database each time a certain process is kicked off and there is never more than one current-run db-- so if the data is lost from the current run it is no big deal. d) It is a Win32 application and PRAGMA page_size = 4096. e) System memory is 512MB and PRAGMA cache_size = 65536 f) Inserts are bundled into one transaction, sometimes as many as 800 at a shot before .Exec(COMMIT) is called. g) No database error codes were generated. h) There are no other programs running when our application is running and Windows Task Manager (processes and performance) shows no unusual activity. Description: Program runs, inserts between 500 and 800 records into the current-run db, and then at various points in reads/inserts (within 10 seconds after the first big insert) the 30-40 second hit will occur. Important note: once that 30-40 second hit is done the program runs great-- for up to 12 hours in fact doing hundreds of thousands of transactions and with great speed. And it only stopped after 12 hours because we completed our tests. We want to buy it but are concerned about the 30-40 second hit. Anyone got an idea what might be going on? Thanks much, Tom Nicholas http://www.thntech.com
