Thread
-
how to trace a backend session
James Pang <jamespang886@gmail.com> — 2024-01-22T07:29:27Z
experts, We migrated our database from Oracle to Postgresql recently, we used to turn on Oracle session trace(that will capture all activities, SQL statements, waiting, waiting time), to do living troubleshooting. could you direct any similar tracing in Postgresql v13 , v14. Thanks, James -
Re: how to trace a backend session
pierre.forstmann@gmail.com — 2024-01-23T18:45:05Z
Hello, I have coded an extension to trace SQL statements for specific backends: https://github.com/pierreforstmann/pg_log_statements (only SQL statements are traced - no wait events data is collected). Pierre Le lun. 22 janv. 2024 à 08:29, James Pang <jamespang886@gmail.com> a écrit : > experts, > We migrated our database from Oracle to Postgresql recently, we used > to turn on Oracle session trace(that will capture all activities, SQL > statements, waiting, waiting time), to do living troubleshooting. could > you direct any similar tracing in Postgresql v13 , v14. > > Thanks, > > James >
-
Re: how to trace a backend session
Dominique Devienne <ddevienne@gmail.com> — 2024-01-24T08:16:57Z
> > Le lun. 22 janv. 2024 à 08:29, James Pang <jamespang886@gmail.com> a > écrit : > >> [...] we used to turn on Oracle session trace(that will capture all >> activities, >> > SQL statements, waiting, waiting time), to do living troubleshooting. >> > could you direct any similar tracing in Postgresql v13 , v14. >> > On Tue, Jan 23, 2024 at 7:45 PM Pierre Forstmann <pierre.forstmann@gmail.com> wrote: > I have coded an extension to trace SQL statements for specific backends: > https://github.com/pierreforstmann/pg_log_statements > (only SQL statements are traced - no wait events data is collected). > There's also the possibility of activating client-side tracing with libpq, if you are using it: https://www.postgresql.org/docs/current/libpq-control.html#LIBPQ-PQTRACE Not the same thing as server-side tracing, of course. But can be useful. --DD PS: Although that trace is not easily machine parse'able, it is at least human readable (kinda...) (see https://www.spinics.net/lists/pgsql/msg218580.html),