RyanMarcus / pg_session_stats

Track PostgreSQL resource usage across a session, including parallelism.

pg_session_stats

This extension lets you track PostgreSQL resource usage at the level of a session (connection). Notably, this library properly tracks usage across parallel workers. To instrument a single query, one should create a new session, execute that query, then end the session.

To install, ensure you have sqlite3 installed and run:

make USE_PGXS=1 install

The extension makes a few assumptions that might not stay true, but are true in PG12:

• That ExecutorEnd hooks are reliably called after all of what could be considered query processing has completed.
• That the /proc/{PID}/cmdline of worker processes will end with the main executor's PID (if anyone knows a better way to connect worker PIDs to the primary executor, let me know).

Since dealing with PostgreSQL's shared memory and locking system is not my idea of a good time, this extention saves the results into a SQLite3 database. You can set the location of this database by setting pg_session_stats.path in your postgresql.conf.

The created DB has a single table called log with the following schema:

CREATE TABLE log (
    master_pid INT,
    my_pid     INT, 
    usage      REAL,
    procstatus TEXT
);

• The master_pid column will contain the PID of the main executor process (i.e., the one given by select pg_backend_pid();).
• The my_pid column contains the PID of executor (which may be equal to master_pid).
• The usage column contains the total amount of CPU time used by that executor.
• The procstatus column contains the output of /proc/{PID}/status for that executor after query processing was complete.