All, I've recently seen two questions on the users' list and a bunch of questions over on StackExchange/ServerFault about mod_jk trying to connect to 0.0.0.0 instead of some sane IP address. Many have comments saying that reverting to mod_jk 1.2.35 seems to fix the issue.
I saw that Mladen recently made a change to jk_util.c (http://svn.apache.org/viewvc?rev=1338797&view=rev) to use memset(0) instead of individually blanking-out the fields in the jk_ws_service_t structure but after checking the patch, it seems that he was very careful to ensure that certain values didn't get set to 0 but some other appropriate value. Also, that change was made after the release of 1.2.36 so it's certainly not the cause. I wonder if anyone can remember some similar change that may, in certain cases, blank-out the server's IP address for mod_jk's workers. There also might be something weird with the lookup of "localhost": on one of the SE questions, the OP did "nslookup localhost" and the request went to a DNS server (which returned NXDOMAIN) and didn't just immediately get 127.0.0.1 from his machine's lookup-service. It seems to be to be pretty stupid for someone to intentionally configure mod_jk to use worker.host=0.0.0.0 (and most have posted configurations indicating otherwise) so maybe there is a weird use-case where host resolves to 0.0.0.0. Any ideas? References: http://markmail.org/message/v2uvl7hgttuje2nf http://serverfault.com/questions/390313/connection-from-apache-to-tomcat-via-mod-jk-not-working http://stackoverflow.com/questions/10663525/apache-tomcat-redirection-error-worker1-connecting-to-backend-failed -chris
signature.asc
Description: OpenPGP digital signature
