strace is a nice tool. Once I figured out that it needed the -f option (for threads maybe as I don't fork other processes) I got good output, and, no, it does not contain any fsync or any other *sync* calls.
So that answers my question. Testing this out on a different Linux box I did not see the cpu and disk I/O patterns which led me to think that a sync was being done, so it must be some issue specific to the first host unrelated to Sqlite. Thanks for your response! Bob ________________________________ From: Richard Hipp <d...@sqlite.org> To: Bob Price <rjp_em...@yahoo.com>; General Discussion of SQLite Database <sqlite-users@sqlite.org> Sent: Wednesday, October 17, 2012 11:32 AM Subject: Re: [sqlite] pragma synchronous=off and journal_mode=wal still syncs to disk On Wed, Oct 17, 2012 at 11:04 AM, Bob Price <rjp_em...@yahoo.com> wrote: It appears that regardless of a "pragma synchronous=off" that there are still syncs to disk done at key points in WAL mode such as in a "pragma wal_checkpoint(RESTART)". I think that this is true based on the application cpu and disk I/O patterns I observe when logging shows the wal checkpoint to be in progress, while running on a Linux host with tons of free memory available for its file cache. > I ran test cases using "strace" and they show no fsyncs occurring when synchronous=off. Do you have a script that shows an fsync (or fdatasync) using strace? >Is this true? Is there any way to configure a connection so that these syncs >do not happen in WAL mode? > >Thanks, >Bob >_______________________________________________ >sqlite-users mailing list >sqlite-users@sqlite.org >http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users > -- D. Richard Hipp d...@sqlite.org _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
