At 08:15 PM 7/14/99 -0600, Alex Rice wrote:
Just wondering if this is what other people are seeing. Frequently (once a day?), ZServer goes out to lunch and I get the following error message:
(102) failure during connect <!-- Connection refused pcgi-wrapper-version 2.0a5 -->
This error means that the pcgi wrapper cannot connect to the pcgi server.
or sometimes, ZServer will completely die and the non-ZServer version of Zope will start instead (what is it called, pcgi publisher?) which of course has the wrong content because I am working on ZServer/ODB3.
You can prevent this by not including information about the pcgi publisher in your PCGI info file. All you really need when using ZServer as your pcgi publisher is the pid file and the socket file directives. You don't need the pcgi publisher and pcgi exe directives.
In my apache error log, this is the last entry:
[Wed Jul 14 19:27:51 1999] [error] [client 206.183.203.132] Premature end of script headers: /home/sites/home/web/zope/cgi-bin/Zope.cgi
All this means is that the pcgi wrapper failed.
I have to manually restart z2.py to get it back. I am starting Z2.py with the -Z option to start the watcher thread, which apparently isn't working or doesn't think ZServer is having problems.
Hmm. When you start ZServer does a zdaemon process start? Do you get any logging messages from zdaemon.
Is this because it's still Alpha? Currently, ZServer dies way too often to use in a production site.
I've found ZServer to be fairly stable. My goal is to have is pretty darn solid by the time Zope 2 is final. Clearly you are having problems, but without more information, it's hard to tell what is responsible.
Here is the command I'm using to start Zope: python z2.py -u zope -Z var/mgr.pid &
Is this the correct usage of the -Z option?
Yes. Note: for debugging purposes, you may want to run ZServer in the foreground so you can see tracebacks that may crop up if it fails.
I'm currently using Zope 2.0a4 on Linux/mips.
Cool. I am very motivated to get ZServer working as well as possible. In your case I need more information about what is failing in order to help you correct your configuration or fix any problems that ZServer may have. Is there any particular pattern to when Zope fails? Is anything written to the ZServer log? Is anything written to the stdout? Does only the PCGI server stop, or do all medusa servers stop? Does the Zope process actually quit, or does it just stop responding? What about the zdaemon process? At this point we may want to take this discussion offline. Also, for everyone out there whose still reading this, I suggest filing bugs in the Collector, in addition to just mailing them to the list :-) Thanks! -Amos