Description of problem: after fresh install of RDO on the box using https://github.com/AthabascaUniversity/openstack-ng things seem to be functional except for the networking: trying to assign external IP to an instance results in error in logs: failed to set SO_REUSE{ADDR|PORT} on DHCP socket: Protocol not available downgrading back to dnsmasq-2.48-13.el6.x86_64 and rebooting machine fixes the issue. Version-Release number of selected component (if applicable): dnsmasq-2.48-14.el6.x86_64 How reproducible: I have been able to reproduce this two times - first time on existing setup, second on fresh install. Steps to Reproduce: 1. either use https://github.com/AthabascaUniversity/openstack-ng or follow http://docs.openstack.org/icehouse/install-guide/install/yum/content/index.html 2. create private network/subnet 3. create public network/subnet 4. create router and configure networks 5. create floating IP 6. create instance 7. try to associate floating IP with instance Actual results: /var/log/neutron/dhcp-agent.log:2014-12-19 10:39:42.214 2938 TRACE neutron.agent.dhcp_agent Stderr: '\ndnsmasq: failed to set SO_REUSE{ADDR|PORT} on DHCP socket: Protocol not available\n' /var/log/messages:Dec 19 10:39:42 foohost dnsmasq[7871]: failed to set SO_REUSE{ADDR|PORT} on DHCP socket: Protocol not available and while nova list report IP as being assigned - it is not available to external connections. Expected results: ability to connect to floater IP Additional info: abovementioned playbook follows closely official OpenStack guide for install on RHEL. redhat-release-server-6Server-6.6.0.2.el6.x86_64
Hi, Is it possible to add steps for reproducing without openstack? like, config for dnsmasq. Options what are used at command line for dnsmasq (or how it is scheduled). network topology. I've discussed with developer, and we are not sure that this is problem in dnsmasq or it is only accident. Thanks&Regards Honza
Jan, I am not sure I, personally can reproduce the problem without OpenStack - as I am not using dnsmasq directly. Invocation parameters would have to be fished out of Quantum code, which I haven't done and not sure if I will be able to do.
I don't think we can do anything until we have a proper reproducer. I also only see a trivial initscript difference between the version reported as working and the version reported as failing.
I believe that this is the very same issue bug #1154953. Same symptoms, same affected version. The above bug report now contains a reproducer and reference to an upstream commit. *** This bug has been marked as a duplicate of bug 1154953 ***