Daan, I've checked the code, the issue is, original code Bharat wrote(generated dnsmasq.conf) cannot deal with VPC. e.g. the VPC router wouldn't listen on eth0 and eth1, but non-vpc router wouldn't listen on eth1 and eth2. Also, some detail configurations of VPC are written into /etc/dnsmasq.d/cloud.conf, rather than dnsmasq.conf, because it need to deal with multiple nics with different range.
So using non-vpc's configuration just break whole VPC function. No chance it can work because eth2(which is internal network interface in VPC) is excluded from listening interface. We need to deal with it ASAP because it's blocking the regression test. In fact I didn't see much solution for now except reverting it since the dnsmasq config command didn't support VPC anyway. BTW, I don't agree that we need dnsmasq configuration file generated by cloudstack, since it's possible that the scripts in the VR would modify the file as well(e.g. in cloud-early-setup), then the modification would be overrided by whole new file generated by management server(which happened in this case). We have to gather them in one place. I've talked with Bharat and he would provide a patch soon to revert the DnsMasqConfigurator, and use bash scripts to do the necessary substitute instead. I think any potential change to dnsmasq config command need to wait for it. --Sheng On Fri, Jul 19, 2013 at 5:54 PM, Sheng Yang <sh...@yasker.org> wrote: > Daan, > > Bharat thinks your fix caused a block bug > https://issues.apache.org/jira/browse/CLOUDSTACK-3589? Could you check it? > > --Sheng > > On Thu, Jul 18, 2013 at 9:52 AM, Bharat Kumar <bharat.ku...@citrix.com>wrote: > >> Hi Dann, >> >> The bug fix >> https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=commitdiff;h=b903262df5e2ff5d174859ce28abae75c4689f0cis >> causing an null value in dnsmasq.config file while creating VPC network >> (bug-id Cloudstack-3589 ). >> >> Can you please take a look at this. >> >> Regards, >> Bharat. > > >
