Yes, we do see some plans logged by the auto_explain. We couldn't find a `auto_explain.log_min_duration_statements` setting but `log_min_duration_statement` as in https://www.postgresql.org/docs/current/runtime-config-logging.html is set to 100 ms.
Unfortunately, due to the amount of traffic we have, we cannot lower the threshold for the auto_explain. `log_lock_waits` is turned on and the logs do not indicate any locks related to the tables in the query. On Fri, 2 Sept 2022 at 05:49, Julien Rouhaud <rjuju...@gmail.com> wrote: > Hi, > > On Thu, Sep 01, 2022 at 08:20:13PM +0100, Matheus Martin wrote: > > We tried running the prepared statement six times as suggested but wasn't > > still able to recreate the original problem. > > > > Perhaps more concerning/relevant is that we have not found any > explanation > > to why the explain plan is not being logged by `auto_explain`. Could this > > be a bug? Shall we report it? > > Just to be sure, do you get at least some plans logged by auto_explain when > queries are executed by the JDBC application? > > Can you try to temporarily lower auto_explain.log_min_duration_statements > to > less than 50ms and see what auto_explain sees for the execution time (and > planning time). > > Another possibility would be some conflicting locks held. If the conflict > happens during the planning auto_explain still won't be triggered as it's > outside the executor. Also, have you enabled log_lock_waits? >
