sqlalchemy apache mod_wsgi与金字塔崩溃

Question

I am running a basic Pyramid application with apache mod_Wsgi. My application setup seems Ok since I can access the home Pyramid template under Apache.

As soon as I issue a request to this app, the apache process crashes with 'premature end of script headers - segmentation fault'. This is how the sqlalchemy code looks for the request. (I've added this code to _ views. _py itself since I am trying to create a RESTful framework without the use of models.)

try:
            print >> sys.stderr, 'entering sql execute'
            value=session.execute(s)
            for row in value:
                    listofclients.append(row[0])
            return listofclients
            session.close()
    except exc.SQLAlchemyError, exc:
            print >> sys.stderr, exc.message

I can see 'entering sql execute' in the apache error log. Immediately after which the process crashes. Commenting out the execute statement and returning sample values works. Running this app under pserve provided by pyramid works like a charm.

Any help why Apache does not like the execute statment? Let me know if you need any other information.

Answer 1

Try reading the mod_wsgi FAQ about process crashes.

• http://code.google.com/p/modwsgi/wiki/FrequentlyAskedQuestions#Apache_Process_Crashes

The most probably cause being:

• http://code.google.com/p/modwsgi/wiki/ApplicationIssues#Python_Simplified_GIL_State_API

To get a better idea use a debugger to capture a stack trace:

• http://code.google.com/p/modwsgi/wiki/DebuggingTechniques#Debugging_Crashes_With_GDB

Answer 2

Following Graham's suggestion and carefully examining the GDB backtrace of the crashed process, it turns out there was a ldap module which was interfering with the cx_Oracle connect string. the ldap_search_str() was the culprit.

Disabling the ldap modules fixed the issue.