Hi, I had a serious problem with my site yesterday. For some unknown reason, Zope seemed to lock up or slow down extremely, so we couldn't get a response anymore. I tried several things: * restart Zope * replace Data.fs by an earlier version * redirect Apache to another Zope server These things didn't work. After a couple of minutes the problem returned. After searching through Apache log files, a colleague found that something weird happened with requests for pages that started with '/error/'. So we changed the rewrite rules to deny those requests, and everything seemed to work again. Since then, we've only had the ususal (read LDAP) problems. My questions: * Were those '/error/'-requests really the cause of the problem? How can I be sure. And why did this happen just now? I've been using those same rewrite rules for over a year now, but I never had this problem before. * Is there a list of things to check for when Zope locks up? Yesterday, we looked at about everything, before we stumbled upon those '/error/'- requests. Could we have found this earlier? * Is there a list of bad requests, ie. requests that lock up Zope. I already denied WebDAV-requests (PROPFIND, LOCK) on Zope's HTTP-port, because of problems with M$ Word and Word-documents. Are there other things that I should deny, or at least handle with care? My configuration: Redhat 8.0 Apache 2.0.40 Zope 2.6.4 CMF 1.2 Bert Vanderbauwhede... -- "I laugh in the face of danger. Then I hide until it goes away." -- Alexander LaVelle Harris
On Thu, May 27, 2004 at 09:49:27AM +0200, Bert Vanderbauwhede wrote:
After searching through Apache log files, a colleague found that something weird happened with requests for pages that started with '/error/'. So we changed the rewrite rules to deny those requests, and everything seemed to work again. Since then, we've only had the ususal (read LDAP) problems.
My questions:
* Were those '/error/'-requests really the cause of the problem? How can I be sure.
Your description is much too vaugue. What is "something weird"? By default Zope contains no object named /error. Unless you have added one, those requests should have simply returned a 404 response.
* Is there a list of things to check for when Zope locks up?
Not really. But Evan's "Debug a Spinning Zope" recipe has proved helpful to me: http://www.zopelabs.com/cookbook/1073504990 I also tend to always turn on "debug logging", aka "big-M logging" (named for the z2.py switch that turned it on in zope 2.6 and lower). When something unusual happens such as very poor performance, running requestprofiler.py on the debug log can be instructive. E.g. you may find certain requests that always hang or take much longer than you expect. This can be a good clue.
we looked at about everything, before we stumbled upon those '/error/'- requests. Could we have found this earlier?
* Is there a list of bad requests, ie. requests that lock up Zope.
No. If there is such as a thing as a "request that locks up zope" it is a bug and should be reported at http://collector.zope.org/Zope . -- Paul Winkler http://www.slinkp.com
On Thu, 2004-05-27 at 10:00, Paul Winkler wrote:
* Is there a list of bad requests, ie. requests that lock up Zope.
No. If there is such as a thing as a "request that locks up zope" it is a bug and should be reported at http://collector.zope.org/Zope .
while 1: pass ?
On Thu, May 27, 2004 at 10:20:10AM -0400, Chris McDonough wrote:
On Thu, 2004-05-27 at 10:00, Paul Winkler wrote:
* Is there a list of bad requests, ie. requests that lock up Zope.
No. If there is such as a thing as a "request that locks up zope" it is a bug and should be reported at http://collector.zope.org/Zope .
while 1: pass
That's not a request. Mornin', Chris. Had your coffee yet? ;-) -- Paul Winkler http://www.slinkp.com
On Thu, 2004-05-27 at 10:32, Paul Winkler wrote:
On Thu, May 27, 2004 at 10:20:10AM -0400, Chris McDonough wrote:
On Thu, 2004-05-27 at 10:00, Paul Winkler wrote:
* Is there a list of bad requests, ie. requests that lock up Zope.
No. If there is such as a thing as a "request that locks up zope" it is a bug and should be reported at http://collector.zope.org/Zope .
while 1: pass
That's not a request.
Mornin', Chris. Had your coffee yet? ;-)
Nope. ;-) But good idea! My point just was that it's definitely possible for an developer to cause Zope to lock up, it doesn't try very hard to protect people from themselves (even if they don't realize they're "developing" anything). Reporting a request that hangs due to something like this to the collector would be sort of pointless. - C
On Thu, May 27, 2004 at 10:35:28AM -0400, Chris McDonough wrote:
On Thu, 2004-05-27 at 10:32, Paul Winkler wrote:
On Thu, May 27, 2004 at 10:20:10AM -0400, Chris McDonough wrote:
On Thu, 2004-05-27 at 10:00, Paul Winkler wrote:
* Is there a list of bad requests, ie. requests that lock up Zope.
No. If there is such as a thing as a "request that locks up zope" it is a bug and should be reported at http://collector.zope.org/Zope .
while 1: pass
That's not a request.
Mornin', Chris. Had your coffee yet? ;-)
Nope. ;-) But good idea!
My point just was that it's definitely possible for an developer to cause Zope to lock up, it doesn't try very hard to protect people from themselves (even if they don't realize they're "developing" anything). Reporting a request that hangs due to something like this to the collector would be sort of pointless.
OK, I see your point. If the request is to in-house code that does something silly, sure, that's user error and not a zope bug. Some of these are a bit obscure to the newbie (e.g. I had an external method happily using urllib2, until a firewall change caused that method to lock up my whole site. I hadn't thought about the lack of a timeout.) I understood the OP to be asking if there are known issues where a fresh-out-of-the-box zope site can be brought down by a certain request. These would go in the collector. -- Paul Winkler http://www.slinkp.com
participants (3)
-
Bert Vanderbauwhede -
Chris McDonough -
Paul Winkler