my file-max is set to 8192, which I thought was plenty. "Normal"
usage shows PG using 2300 handles, nsd using about 1000, and
everything else using 500. However, twice today PG was restarted by
my watchdog script after stuff started failing with "file table
overflow" all over my logs. /proc/sys/fs/file-nr confirms that it
did indeed max out. Should I just add to the limit, or is this
indicative of a PG config problem or bug?