Re: [Zope] (116) unable to connect, fd=3
Michel Pelletier wrote:
I wish I knew more of its nature. :-/
Put just this bit of DTML into a method (lets it's called 'themethod') somewhere in your root Zope folder.
*Shut down Zope*
go to 'Zopedir/lib/python'
do this:
[michel@aldous python]$ python Python 1.5.1 (#2, Jan 26 1999, 10:47:10) [GCC 2.7.2.3] on linux2 Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam
import ZPublisher, Main
ZPublisher.Main('themethod?pin=blah&from=blah@blah.com&message=blahblah' )
'themethod' should be the name of the method containing the errant DTML code
'pin', 'from', and 'message' should be sensable values that make your code fail through the web also.
ZPublisher.Main has the effect of calling 'themethod' just like a web request would. Something is happening to make your pcgi connection fail with the Zope long running process, and thus you cannot see the error that the python process is giving you. Since pcgi is not involved in this little test, you will see the error printed on your terminal. It could be a syntax error, a core dump, an SMTP error that causes something to get sent to stderr instead of raising an exception, who knows. Sniff it out.
Well, first thing this morning, I tried it out, and could not reproduce the error. In fact it successfully paged the pin number I used. Here is what I have found: If I start ZPublisher manually within python (import ZPublisher, Main) the page works, if it is started from a connection through the web (Zope.cgi) I get the (116) unable to connect, fd=3 error. Are we getting warmer? -- Tim Hawes tim.hawes@ncmail.net
Tim Hawes wrote:
If I start ZPublisher manually within python (import ZPublisher, Main) the page works, if it is started from a connection through the web (Zope.cgi) I get the (116) unable to connect, fd=3 error. Are we getting warmer?
Tim, Good sleuthing. The current version of pcgi attempts to launch the LRP and connect via dead reckoning. Digital Creations agrees with (even encourages) the idea that process launch should be moved out of pcgi. We have even added an out-of-band byte to the protocol to facilitate LRP <-> pcgi negotiation. Question: When you started ZPublisher manually, did you get any spewed output to stdout/stderr? Best regards, Jeff Bauer Rubicon, Inc.
Jeff Bauer wrote:
Tim Hawes wrote:
If I start ZPublisher manually within python (import ZPublisher, Main) the page works, if it is started from a connection through the web (Zope.cgi) I get the (116) unable to connect, fd=3 error. Are we getting warmer?
Tim,
Good sleuthing. The current version of pcgi attempts to launch the LRP and connect via dead reckoning. Digital Creations agrees with (even encourages) the idea that process launch should be moved out of pcgi. We have even added an out-of-band byte to the protocol to facilitate LRP <-> pcgi negotiation.
Question: When you started ZPublisher manually, did you get any spewed output to stdout/stderr?
Best regards,
Jeff Bauer Rubicon, Inc.
_______________________________________________ Zope maillist - Zope@zope.org http://www.zope.org/mailman/listinfo/zope
(For developer-specific issues, use the companion list, zope-dev@zope.org - http://www.zope.org/mailman/listinfo/zope-dev )
Update on this problem: I copied the database file to another instance of Zope (same version) on the same machine. pcgi returned a corrupt database error (the other instance did not). I truncated the database to the the block with the reported problem, and I no longer get this error when pcgi is started with Zope.cgi. I recompiled the initial instance (with the latest ZPublisher module) and copied the truncated database back, and all seems well. Hmmm. . . . I am becoming more eager to see the database management stuff promised in the next release.
participants (2)
-
Jeff Bauer -
Tim Hawes