Hello Vincent On Wednesday 22 February 2012 03:39:38 Vincent Lefevre wrote: > I'll try to do some tests when I have the time. But there's > no reason eth0 would be IPv6 only here (unless there's a huge > bug in the kernel). I'm prety sure your apache tries to start before dhclient has given the server its IPv4. And you'll have the local link IPv6 address already. The patch of libapr1 will fix your hostname problem: https://issues.apache.org/bugzilla/show_bug.cgi?id=52709 Basically, all loopback IPv4 resolution is broken while dhclient is running the first time, that is when you have IPv6 on eth0 but no IPv4 on it yet. Then libapr1 internal function call_resolver is wrongly using AI_ADDRCONFIG so that loopback IPv4 addresses are ignored. That is why a manual restart (after dhclient completed) works. I did understand your apache starts. Because you probably "Listen 80" rather than "Listen 127.0.1.1:80". 127.0.1.1 resolution would fail then, just like the resolution of ypig in your case. I tested it yesterday. Believe me, the source of your problems are in libapr1, not in apache2 packages, nor in the kernel. Je te parie une bierre si tu veux! ;)
Attachment:
signature.asc
Description: This is a digitally signed message part.