Sebastian Lühnsdorf wrote:
hi!
we are still having problems with freezing zope servers. we are relaying requests to an http server in the intranet, that means we are retrieving the page via httplib. under some circumstances, i.e. with slow pages and several concurrent requests, Zope freezes completely (all threads) until the requested page is fully processed. If somebody has experienced likewise and/or has any info or help.
The cause of the freeze should be that all of zope's worker threads are, well, working. I guess you know that since you increased that value in the config. There may be other tricks you can play, but here's what I can tell you: - Use timeoutsocket.py (-> google) to pose a timeout on hanging requests. This is only usefull if your external server is really slow/hangs and you can afford not to display you external pages if that happens. - Modify you product/external method to cache the pages, see for instance the zope product KebasData how to do that. - retrieve the external pages through a proxycache like squid. - Let an external (to zope) job retrieve the pages regularly and publish them through something like LocalFS (there's a newer product, just forgot it's name). HTH, oliver