[postgis-tickets] [PostGIS] #4857: Large ST_Union makes PostgreSQL temporarily unrestartable

[PostGIS]

#4857: Large ST_Union makes PostgreSQL temporarily unrestartable
-----------------------+---------------------------
  Reporter:  grassick  |      Owner:  pramsey
      Type:  defect    |     Status:  new
  Priority:  high      |  Milestone:  PostGIS 3.1.2
 Component:  postgis   |    Version:  3.1.x
Resolution:            |   Keywords:
-----------------------+---------------------------

Comment (by pramsey):

 What you're seeing is inconsistent with what we know from documentation
 and from experience.

 * You see a swift cancel from PgAdmin, which is causing a SIGINT to be
 sent to the backend and which PostGIS / GEOS handle with alacrity.
 * You are seeing a slow cancel from the shutdown.
 * According to the documentation, and ordinary shutdown sends SIGTERM, and
 PostGIS/GEOS ignore that signal (as the documentation would indicate they
 *should*, since during a normal shutdown PgSQL is expected to let running
 queries run to completion).
 * You show a log that *says* it is running a "fast shutdown" which the
 documentation *says* should generate a SIGINT which we know from above
 generates a swift cancel, and yet you observe takes longer than you'd
 expect.

 The only possibility I can see that matches all the above is that the
 "fast shutdown" does not in fact immediately send a SIGINT, but first
 sends a SIGTERM but waits for a bit and then sends SIGINT if things
 haven't ended on their own.

 That still doesn't leave anything for us to fix. We are doing everything
 *right*. In the event of a SIGINT, we swiftly stop. In the event of a
 SIGTERM, we just let things run. I don't see how this should remain an
 open PostGIS ticket.

-- 
Ticket URL: <https://trac.osgeo.org/postgis/ticket/4857#comment:15>
PostGIS <http://trac.osgeo.org/postgis/>
The PostGIS Trac is used for bug, enhancement & task tracking, a user and developer wiki, and a view into the subversion code repository of PostGIS project.