Can you re-run your strace as "strace -tt" and look at the timings to help pinpoint it?
Michael D. Black Senior Scientist Advanced Analytics Directorate Advanced GEOINT Solutions Operating Unit Northrop Grumman Information Systems ________________________________________ From: [email protected] [[email protected]] on behalf of Keith Chew [[email protected]] Sent: Tuesday, December 04, 2012 5:00 PM To: General Discussion of SQLite Database Subject: EXT :Re: [sqlite] Tracing latencies On Wed, Dec 5, 2012 at 11:10 AM, Keith Chew <[email protected]> wrote: > The strange thing is that I am setting sqlite it to use WAL, > autocheckpoint off and synchronous off. Even in this setup, I still > see > 350ms transactions times for less than 3 TPS. A bit hard to > believe, so I am now doing a strace to find out exactly what is > hitting the disk. II can confirm using strace that there are no fsyncs happening from the application (which we expect because synchronous=0). So, it must be something else that is causing these blocks. Somekind of file locking issue (particular to my environment)? What else should I be looking at? Regards Keith _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
