crashes, viability, etc
I have to restart my zope server every 1-3 days; no clear reason for it, as it dies silently, simply failing to respond on port 8080. I'm running pretty vanilla, with a virtual host monster, very little content, and ZUBB installed with a grand total of one discussion area. I'm on solaris 2.8, if that matters at all. This kind of behavior isn't making a great case to me to recommend the use of zope commercially -- the only time I've come across zope commercially, so far, has been with a customer who (anecdotally, as a coworker is the one who deals with them) has regular problems with zope threads and so forth. That, I can chalk up to bad code... but in my case, with very little code in place, I can't see a good reason for the crashes (or dropping port 8080's, at least, which amounts to the same thing ... I haven't had much time to diagnose, as I'm often in a rush to get the server back running again) to be happening with that kind of frequency. I've seen it happen under pretty variable circumstances, too -- using the /manage interface to upload a file, for example, ended up with an non-running zope (nevermind that the file appeared to upload fine first, before the crashing, as it was there on restart). There's also been some that seemed to simply result from loading a page, or posting a message to the one zubb topic I've got (though I recognize that may be a subset of the file upload situation). This is disappointing, because I really like Zope's architectural notions. The philosophy is great, especially when I compare it with the more "procedural" mindset of coding in Cold Fusion, which is what I _do_ do, professionally. At this point, though, I'd rate CF as more robust -- sure, it crashes on Solaris occasionally (in fact, I'm running zope and cf side-by-side for comparison right now), but that I _have_ tracked down to bad code usually, and when it does, it often has the courtesy (out of the box) to restart itself. But I'd really prefer a more zope-style object-orientedism. That said, I'm looking for a few things, here: 1. Pointers as to how to diagnose an intermittent problem like mystery crashes; 2. Pointers as to whether or not someone has developed a similar auto-restart feature (see what I say about CF, above) for zope; 3. Anecdotes to counter the ones I'm getting from the workplace, and my own experience, on the subject of stability. Anything folks can offer on any or all of the above topics will be met with pleased interest by me; I'm here to learn. Thanks -- Fred Hicks <iago@iago.net>
Iago wrote:
That said, I'm looking for a few things, here:
1. Pointers as to how to diagnose an intermittent problem like mystery crashes;
I have a HOWTO on the subject -- http://www.zope.org/Members/matt/StabilityHOWTO Note that there is a known problem with Python 2.1.2 and earlier which affects Zope stability; in particular, the code path traversed by Zope when compiling restricted python expressions will trigger the bug. This will be fixed in Python 2.1.3. -- Matt Kromer Zope Corporation http://www.zope.com/
Iago writes:
I have to restart my zope server every 1-3 days; no clear reason for it, as it dies silently, simply failing to respond on port 8080. I have seen this problem since ages on Solaris (2.5, 2.6) starting with Zope 2.1.6 and Python 1.5.2.
It occured only very rarely. Interestingly, the frequency drastically went down when we checked Zope's livelyness every 5 minutes. Apparently, Zope does not like to be lazy ;-) It happened too raraly that I had a chance to analyse the problem. Thus, I can not suggest a solution to you. Dieter
participants (3)
-
Dieter Maurer -
Iago -
Matthew T. Kromer