Zope hangs, with processes being idle ... (rather urgent problem)
Hi! I have a Zope 2.5.1 using ZEO on Linux, with Apache as proxy server. The two ZEO clients run fine for a couple of hours, and then suddenly one of them stops responding. It's the type of Zope hangs when all processes are fine, but ZServer just doesn't respond (load is very low, with the CPU 90% idle, so it is not a performance issue or recursion problem). There is no error before this situation happens. The processes just don't respond any more. The second ZEO keeps running, but that machine also hangs from time to time, so I guess ZEO is not the problem, and the hardware or any Linux-specific problems aren't either. When the ZServer hangs, I can not access it directly, so it is no problem with the Apache. You see, I don't have any clue on what is happening here. The servers worked fine for months, and I don't know what specific change could have caused the problem. How can such a situation be debugged? Cheers Joachim
Joachim Werner writes:
I have a Zope 2.5.1 using ZEO on Linux, with Apache as proxy server. The two ZEO clients run fine for a couple of hours, and then suddenly one of them stops responding. It's the type of Zope hangs when all processes are fine, but ZServer just doesn't respond (load is very low, with the CPU 90% idle, so it is not a performance issue or recursion problem).
There is no error before this situation happens. The processes just don't respond any more. The second ZEO keeps running, but that machine also hangs from time to time, so I guess ZEO is not the problem, and the hardware or any Linux-specific problems aren't either.
When the ZServer hangs, I can not access it directly, so it is no problem with the Apache.
You see, I don't have any clue on what is happening here. The servers worked fine for months, and I don't know what specific change could have caused the problem.
How can such a situation be debugged? Maybe, you use the "-M" logging and analyse the log with Chris' "requestprofiler". You may see unfinished requests.
Dieter
I have a Zope 2.5.1 using ZEO on Linux, with Apache as proxy server. The two ZEO clients run fine for a couple of hours, and then suddenly one of them stops responding. It's the type of Zope hangs when all processes are fine, but ZServer just doesn't respond (load is very low, with the CPU 90% idle, so it is not a performance issue or recursion problem).
There is no error before this situation happens. The processes just don't respond any more. The second ZEO keeps running, but that machine also hangs from time to time, so I guess ZEO is not the problem, and the hardware or any Linux-specific problems aren't either.
When the ZServer hangs, I can not access it directly, so it is no problem with the Apache.
You see, I don't have any clue on what is happening here. The servers worked fine for months, and I don't know what specific change could have caused the problem.
How can such a situation be debugged?
Dieter Maurer wrote:
Maybe, you use the "-M" logging and analyse the log with Chris' "requestprofiler". You may see unfinished requests.
Thanks! That's what I did. I think I found the problem: I have some Objects that inherit from Image, and their "data" attribute (that's where the image data is held) wasn't initialized correctly in some cases. So whenever somebody tried to display one of those objects, Zope started streaming back the image data, which made the request hang when the data attribute was broken. Cheers Joachim
participants (2)
-
Dieter Maurer -
Joachim Werner