IE problems with Zope Solaris binary?
Hello, I'm running Zope-2.3.2-solaris-2.6-sparc binary release on a Sparc Ultra 60 running Solaris 2.8. I don't have root privilege on my machine, and gcc compiler is not installed, so I just downloaded the Solaris binary. When I start Zope in debug mode, I see: ZServer PCGI Server started .. Unix socket: /home/barg/Zope-2.3.2-solaris-2.6-sparc/var/pcgi.soc I connect directly to my zope site using this URL: http://nickel.as.arizona.edu:8080 Since the binary version is using PCGI, would there be any advantage to starting Zope with the "-p" option and going through my public_html directory? Everything appeared to be working fine until I started investigating a problem that I believed was related to an AOL Proxy server caching, but now I don't know. I have a set of forms that allow users to edit a MySQL database (MySQL 3.23.38, MySQL-python-0.3.5, ZMySQLDA-2.0.6). I develop on a linux system, running Zope 2.3.2 compiled from source. Then I deploy on my Sun. Users accessing my forms from Windows machines running either IE or a AOL enabled browser can query my database, but when they actually try to 'Submit' an update, they get the infamous IE page saying: This page cannot be displayed. The page your are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or your may need to adjust your browser settings. ..... Now, what "should" happen is, when they press the 'Submit' button, a DTML method is called (using method="POST"), this method calls some Python scripts, a ZSQL method, then sends a reply to the client's browser. Of course this never happens on a PC running IE or Nescape 4.7. I do not see this behavior if I come into my Zope site from another Sun workstation in my office (using Netscape 4.7), but I can go down the hall and get the same error on a PC running Windows 2000 and IE. Can anyone tell me what may be going on, or how I can go about debugging this? Thanks you, --irene Irene Barg Email: ibarg@as.arizona.edu Steward Observatory Phone: 520-621-2602 University of Arizona FAX: 520-621-1891 Tucson, AZ 85721 http://nickel.as.arizona.edu/~barg
Irene Barg writes:
.... but when they actually try to 'Submit' an
update, they get the infamous IE page saying:
This page cannot be displayed. The page your are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or your may need to adjust
your browser settings. .....
Now, what "should" happen is, when they press the 'Submit' button, a DTML method is called (using method="POST"), this method calls some Python scripts, a ZSQL method, then sends a reply to the client's browser. Of course this never happens on a PC running IE or Nescape 4.7. I do not see this behavior if I come into my Zope site from another Sun workstation in my office (using Netscape 4.7), but I can go down the hall and get the same error on a PC running Windows 2000 and IE. Can anyone tell me what may be going on, or how I can go about debugging this? Shane's TCPWatch utility (or any other TCPLogger) may help you to see precisely what happens and what the difference between the two constellations is.
BTW, does your Zope die when you see this error (you should have enabled logging with "STUPID_LOG_FILE=logfile" and check the logfile). If so, you may have a core dump or an error indication in the log file. Dieter
Dieter, I can't try TCPWatch or webdebug because they both require Python threads. I don't have root privilge on this system and the Python version running is 1.5.2 without threads. I had thought that the Python library that came with Zope was self-contained? When I downloaded the Zope Solaris binary, I placed the required Zope modules I needed in my PYTHONPATH directory, then edited the Zope "start" script to read: nickel% pwd /home/barg/Zope-2.3.2-solaris-2.6-sparc nickel% more start #! /bin/sh PYTHONPATH=.:/home/barg/zope/lib/python1.5/site-packages:/home/barg/pymodules export PYTHONPATH LD_LIBRARY_PATH=.:/home/barg/mysql/lib/mysql export LD_LIBRARY_PATH reldir=`dirname $0` PYTHONHOME=`cd $reldir; pwd` export PYTHONHOME exec /home/barg/Zope-2.3.2-solaris-2.6-sparc/bin/python \ $PYTHONHOME/z2.py \ -u barg "$@" Assuming that the Python built with Zope had threads. Per your suggestion, I restarted it with STUPID_LOG_FILE and with debugging turned on. Re-issued the same Update request and noticed that I now have a 'core' file with the same time stamp as my last request, but the Zope.log doesn't give me any more info than that at startup: nickel% tail var/Zope.log nickel% tail var/Zope.log 2001-08-06T21:41:34 INFO(0) ZServer HTTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8080 ------ 2001-08-06T21:41:34 INFO(0) ZServer FTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8021 ------ 2001-08-06T21:41:34 INFO(0) ZServer PCGI Server started at Mon Aug 6 14:41:34 2001 Unix socket: /home/barg/Zope-2.3.2-solaris-2.6-sparc/var/pcgi.soc BTW, I develop on a my Toshiba notebook, running Linux, Python 2.0 and Zope 2.3.2 built from source. I do not see this problem using it as the Zope server, coming from IE on the PC. I guess I'll have to install my own version of Python (with threads), and try to compile Zope on my Solaris system. This concerns me a little because the gcc compiler is not on my Sun system either. oh well :-( I appreciate your input. Cheers, irene barg .. Dieter Maurer wrote:
Irene Barg writes:
.... but when they actually try to 'Submit' an
update, they get the infamous IE page saying:
This page cannot be displayed. The page your are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or your may need to adjust
your browser settings. .....
Now, what "should" happen is, when they press the 'Submit' button, a DTML method is called (using method="POST"), this method calls some Python scripts, a ZSQL method, then sends a reply to the client's browser. Of course this never happens on a PC running IE or Nescape 4.7. I do not see this behavior if I come into my Zope site from another Sun workstation in my office (using Netscape 4.7), but I can go down the hall and get the same error on a PC running Windows 2000 and IE. Can anyone tell me what may be going on, or how I can go about debugging this? Shane's TCPWatch utility (or any other TCPLogger) may help you to see precisely what happens and what the difference between the two constellations is.
BTW, does your Zope die when you see this error (you should have enabled logging with "STUPID_LOG_FILE=logfile" and check the logfile). If so, you may have a core dump or an error indication in the log file.
Dieter
-- ---------------------------------------------------------------- Irene Barg Email: ibarg@as.arizona.edu Steward Observatory Phone: 520-621-2602 933 N. Cherry Ave. University of Arizona FAX: 520-621-1891 Tucson, AZ 85721 http://nickel.as.arizona.edu/~barg ----------------------------------------------------------------
Hi Irene, you can also debug on client side with ethereal or tcpwatch or whatever. Ethereal would be the big hammer buts very powerful. (http://www.ethereal.com, make sure you download libcap for windows too. You usually capture with "port 80" as filter expression, then "follow tcp stream" on the packets). It might be the problem if your windows clients use excessive HTTP1/1 in opposition to some old client on the other solaris box. This can be a problem with threads. I think installation from sources should fix that. Regards Tino --On Montag, 6. August 2001 14:57 -0700 Irene Barg <ibarg@as.arizona.edu> wrote:
Dieter,
I can't try TCPWatch or webdebug because they both require Python threads. I don't have root privilge on this system and the Python version running is 1.5.2 without threads. I had thought that the Python library that came with Zope was self-contained?
When I downloaded the Zope Solaris binary, I placed the required Zope modules I needed in my PYTHONPATH directory, then edited the Zope "start" script to read:
nickel% pwd /home/barg/Zope-2.3.2-solaris-2.6-sparc nickel% more start # ! /bin/sh PYTHONPATH=.:/home/barg/zope/lib/python1.5/site-packages:/home/barg/pymodules export PYTHONPATH LD_LIBRARY_PATH=.:/home/barg/mysql/lib/mysql export LD_LIBRARY_PATH reldir=`dirname $0` PYTHONHOME=`cd $reldir; pwd` export PYTHONHOME exec /home/barg/Zope-2.3.2-solaris-2.6-sparc/bin/python \ $PYTHONHOME/z2.py \ -u barg "$@"
Assuming that the Python built with Zope had threads. Per your suggestion, I restarted it with STUPID_LOG_FILE and with debugging turned on. Re-issued the same Update request and noticed that I now have a 'core' file with the same time stamp as my last request, but the Zope.log doesn't give me any more info than that at startup:
nickel% tail var/Zope.log
nickel% tail var/Zope.log 2001-08-06T21:41:34 INFO(0) ZServer HTTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8080 ------ 2001-08-06T21:41:34 INFO(0) ZServer FTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8021 ------ 2001-08-06T21:41:34 INFO(0) ZServer PCGI Server started at Mon Aug 6 14:41:34 2001 Unix socket: /home/barg/Zope-2.3.2-solaris-2.6-sparc/var/pcgi.soc
BTW, I develop on a my Toshiba notebook, running Linux, Python 2.0 and Zope 2.3.2 built from source. I do not see this problem using it as the Zope server, coming from IE on the PC. I guess I'll have to install my own version of Python (with threads), and try to compile Zope on my Solaris system. This concerns me a little because the gcc compiler is not on my Sun system either. oh well :-(
I appreciate your input. Cheers, irene barg
..
Dieter Maurer wrote:
Irene Barg writes:
.... but when they actually try to 'Submit' an
update, they get the infamous IE page saying:
This page cannot be displayed. The page your are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or your may need to adjust
your browser settings. .....
Now, what "should" happen is, when they press the 'Submit' button, a DTML method is called (using method="POST"), this method calls some Python scripts, a ZSQL method, then sends a reply to the client's browser. Of course this never happens on a PC running IE or Nescape 4.7. I do not see this behavior if I come into my Zope site from another Sun workstation in my office (using Netscape 4.7), but I can go down the hall and get the same error on a PC running Windows 2000 and IE. Can anyone tell me what may be going on, or how I can go about debugging this? Shane's TCPWatch utility (or any other TCPLogger) may help you to see precisely what happens and what the difference between the two constellations is.
BTW, does your Zope die when you see this error (you should have enabled logging with "STUPID_LOG_FILE=logfile" and check the logfile). If so, you may have a core dump or an error indication in the log file.
Dieter
-- ---------------------------------------------------------------- Irene Barg Email: ibarg@as.arizona.edu Steward Observatory Phone: 520-621-2602 933 N. Cherry Ave. University of Arizona FAX: 520-621-1891 Tucson, AZ 85721 http://nickel.as.arizona.edu/~barg ----------------------------------------------------------------
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
Tino, Dieter, I installed Python 2.0 (with threads) from source, plus Zope-2.3.3-src, and I get the same symptom. Ok, now I really want to know what's going on. Since the Zope compile gave me a lot of warnings with Python 2.0, I decided to re-start the Solaris binary version of Zope, and run tcpwatch.py, i.e: python tcpwatch.py 8086 nickel.as.arizona.edu 8080 Go to my IE browser on a PC, enter the URL: http://nickel.as.arizona.edu:8086/Phillips Tcpwatch logs each hit with a timestamp in the left window, but nothing appears the right window. I can query the database, but when I go to do an Update (which uses method="POST"), tcpwatch dies with the following message (note, I tried with both Python2.0 and Python2.1, they both give the same msg): ****output from tcpwatch.py error: ***** python tcpwatch.py 8086 nickel.as.arizona.edu 8080 error: uncaptured python exception, closing channel <proxy_server listening :8086 at 236ff4> (socket.error:(146, 'Connection refused') [/home/barg/local/lib/python2.1/asyncore.py|poll|95] [/home/barg/local/lib/python2.1/asyncore.py|handle_read_event|377] [tcpwatch.py|handle_accept|67] [tcpwatch.py|connectionFactory|301] [tcpwatch.py|__init__|229] [tcpwatch.py|__init__|113] [tcpwatch.py|__init__|81] [/home/barg/local/lib/python2.1/asyncore.py|connect|308]) error: uncaptured python exception, closing channel <ProxyConnection connected at 2376fc> (exceptions.AttributeError:buffer [/home/barg/local/lib/python2.1/asyncore.py|poll|95] [/home/barg/local/lib/python2.1/asyncore.py|handle_read_event|383] [/home/barg/local/lib/python2.1/asynchat.py|handle_read|99] [tcpwatch.py|collect_incoming_data|120] [/home/barg/local/lib/python2.1/asyncore.py|__getattr__|358]) And, the var/Zope.log looks like this: 2001-08-09T15:44:19 ERROR(200) zdaemon zdaemon: Thu Aug 9 08:44:19 2001: Aiieee! 10318 exited with error code: 138 ------ 2001-08-09T15:44:19 INFO(0) zdaemon zdaemon: Thu Aug 9 08:44:19 2001: Houston, we have forked So, zope dies, dumps core, and restarts. My quess tcpwatch dies when zope did (thus the error 'Connection Refused'). I can only guess that this has something to do with me running zope as user 'barg'. It doesn't matter, I've given up, I'm going to re-write the zope pages in just plain CGI. Your help has been greatly appreciated! --irene barg Tino Wildenhain wrote:
Hi Irene,
you can also debug on client side with ethereal or tcpwatch or whatever. Ethereal would be the big hammer buts very powerful. (http://www.ethereal.com, make sure you download libcap for windows too. You usually capture with "port 80" as filter expression, then "follow tcp stream" on the packets).
It might be the problem if your windows clients use excessive HTTP1/1 in opposition to some old client on the other solaris box.
This can be a problem with threads. I think installation from sources should fix that.
Regards Tino
--On Montag, 6. August 2001 14:57 -0700 Irene Barg <ibarg@as.arizona.edu> wrote:
Dieter,
I can't try TCPWatch or webdebug because they both require Python threads. I don't have root privilge on this system and the Python version running is 1.5.2 without threads. I had thought that the Python library that came with Zope was self-contained?
When I downloaded the Zope Solaris binary, I placed the required Zope modules I needed in my PYTHONPATH directory, then edited the Zope "start" script to read:
nickel% pwd /home/barg/Zope-2.3.2-solaris-2.6-sparc nickel% more start # ! /bin/sh PYTHONPATH=.:/home/barg/zope/lib/python1.5/site-packages:/home/barg/pymodules export PYTHONPATH LD_LIBRARY_PATH=.:/home/barg/mysql/lib/mysql export LD_LIBRARY_PATH reldir=`dirname $0` PYTHONHOME=`cd $reldir; pwd` export PYTHONHOME exec /home/barg/Zope-2.3.2-solaris-2.6-sparc/bin/python \ $PYTHONHOME/z2.py \ -u barg "$@"
Assuming that the Python built with Zope had threads. Per your suggestion, I restarted it with STUPID_LOG_FILE and with debugging turned on. Re-issued the same Update request and noticed that I now have a 'core' file with the same time stamp as my last request, but the Zope.log doesn't give me any more info than that at startup:
nickel% tail var/Zope.log
nickel% tail var/Zope.log 2001-08-06T21:41:34 INFO(0) ZServer HTTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8080 ------ 2001-08-06T21:41:34 INFO(0) ZServer FTP server started at Mon Aug 6 14:41:34 2001 Hostname: nickel.as.arizona.edu Port: 8021 ------ 2001-08-06T21:41:34 INFO(0) ZServer PCGI Server started at Mon Aug 6 14:41:34 2001 Unix socket: /home/barg/Zope-2.3.2-solaris-2.6-sparc/var/pcgi.soc
BTW, I develop on a my Toshiba notebook, running Linux, Python 2.0 and Zope 2.3.2 built from source. I do not see this problem using it as the Zope server, coming from IE on the PC. I guess I'll have to install my own version of Python (with threads), and try to compile Zope on my Solaris system. This concerns me a little because the gcc compiler is not on my Sun system either. oh well :-(
I appreciate your input. Cheers, irene barg
..
Dieter Maurer wrote:
Irene Barg writes:
.... but when they actually try to 'Submit' an
update, they get the infamous IE page saying:
This page cannot be displayed. The page your are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or your may need to adjust
your browser settings. .....
Now, what "should" happen is, when they press the 'Submit' button, a DTML method is called (using method="POST"), this method calls some Python scripts, a ZSQL method, then sends a reply to the client's browser. Of course this never happens on a PC running IE or Nescape 4.7. I do not see this behavior if I come into my Zope site from another Sun workstation in my office (using Netscape 4.7), but I can go down the hall and get the same error on a PC running Windows 2000 and IE. Can anyone tell me what may be going on, or how I can go about debugging this? Shane's TCPWatch utility (or any other TCPLogger) may help you to see precisely what happens and what the difference between the two constellations is.
BTW, does your Zope die when you see this error (you should have enabled logging with "STUPID_LOG_FILE=logfile" and check the logfile). If so, you may have a core dump or an error indication in the log file.
Dieter
-- ---------------------------------------------------------------- Irene Barg Email: ibarg@as.arizona.edu Steward Observatory Phone: 520-621-2602 933 N. Cherry Ave. University of Arizona FAX: 520-621-1891 Tucson, AZ 85721 http://nickel.as.arizona.edu/~barg ----------------------------------------------------------------
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
-- ---------------------------------------------------------------- Irene Barg Email: ibarg@as.arizona.edu Steward Observatory Phone: 520-621-2602 933 N. Cherry Ave. University of Arizona FAX: 520-621-1891 Tucson, AZ 85721 http://nickel.as.arizona.edu/~barg ----------------------------------------------------------------
Irene Barg writes:
Tcpwatch logs each hit with a timestamp in the left window, but nothing appears the right window. You must click on an entry in the left window to view its details in the right one....
.... So, zope dies, dumps core, and restarts. My quess tcpwatch dies when zope did (thus the error 'Connection Refused'). You are right!
I can only guess that this has something to do with me running zope as user 'barg'. It doesn't matter, I've given up, I'm going to re-write the zope pages in just plain CGI. Sad!
Dieter
participants (3)
-
Dieter Maurer -
Irene Barg -
Tino Wildenhain