Re: Unable to connect. Log : PDTWARNING: worker took too long to start; canceled

Linus Kamb escribió:

> Hi,
>
> Every morning for the past several days, when I come in to work, I find the
> postgres process running at 100% CPU and the logs filled with entries every
> minute with the warning "PDTWARNING:  worker took too long to start;
> canceled"
>
> This appears to be almost exactly as this other thread:
>
> http://www.postgresql.org/message-id/AANLkTimt7_wbadtOkM9HqzeHmADmxb2TvuQD5yBUwkdZ@mail.gmail.com

I noticed that the thread was missing a couple of emails in the
archives, and that I had them in my mbox.  I have just forwarded them
("bounced" them in mutt's terms) to the list and I think the archive is
complete now for that thread; I apologize to subscribers that just
received a couple of very old emails.

Anyway, that thread died without finding the root cause of the problem,
but Tom suspected that postmaster was being kept busy doing something
that precluded it from running its usual tasks.  For the OP of the other
thread, stracing the postmaster seemed to bring it back to normal life.

Would it be possible to attach to the postmaster with GDB and get a
stack trace when it's stuck?  That might be helpful; or, if the mere act
of tracing it brings it back to life, it might not -- but instead it
would make me think it's a platform bug of some sort.

If you remove the Glassfish thing, does the problem go away?

(Note: the PDTWARNING thing happens because you don't have a white space
at the end of your log_line_prefix.  Adding one does wonders for
readability.)

--
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services