interpreter crashes

[Paul Rubin]

Peter Hansen <peter at engcorp.com> writes:
> > How frequently does this type of thing happen?  If Python is in the habit
> > of crashing randomly, that may make it inadvisable to write
> > long-running servers that need to stay up.  Perhaps it's more
> > appropriate to write servers Apache-style, forking new processes and
> > reinitializing every so often.
> 
> Not necessary if your system is stable.  Python can run indefinitely
> unless you have hit some obscure bug.  As a small trivial example 
> of its stability, we have had our Intranet server running
> under Zope (on Redhat 7.0) for the last 84 days straight, being
> used daily and extensively by at least twenty people.  No troubles.

This is encouraging to hear.  However, obscure bugs do exist.  I may
be hitting a bug in some extension module or in something having to do
with module reloading.  

> We are also developing applications based on Python which will
> be expected to run potentially for years without being shut down ...
> and I have no concerns about this (at least none related to Python! ;-).

This is my concern--if they're EXPECTED to run for years without
shutdown, but they crash unexpectedly instead, that can be a big
problem.  Best is to not have them crash, but next-best is to plan
ahead for occasional crashes.  So I'm wondering what I need to do.

Note that if you're planning to run a server for years without
attention, besides UPS's you'll also want to have some kind of
hardware watchdog timer on the hosting computer--another form of crash
tolerance.