DHCP broken for Openstack Nova instances since kernel v3.3
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
linux (Ubuntu) |
Triaged
|
High
|
Unassigned |
Bug Description
When configured to use FlatDHCPManager in multi-host, Openstack Nova (Folsom version) instances no longer get their IP address via DHCP from the server running local on the compute node.
For those who are not familiar with Openstack configurations, the compute nodes' VMs come up in a typical bridged networking configuration, similar to the default libvirt configuration, ie:
Public interface (public network): eth0 (192.168.20.11/24)
Private interface (vm network): eth1 /w br100 (10.0.0.3/24)
A dnsmasq process is running bound to 10.0.0.3 and serves DHCP and DNS to the instances on this node based on a config file that is updated with each new instance.
In the case of this bug, instances never successfully get their IP. Doing some packet dumping on the compute node, it looks like VMs successfully spawn and boadcast a DHCP DISCOVER. The local DHCP server receives the discover and sends an OFFER. The OFFER either never makes it to the instance, or the REQUEST for the offered IP never makes it back to the server on the compute node. I've yet to backdoor into an instance and do any snooping from within the VM.
I first noticed this when testing Folsom on Quantal. The same version of Nova works fine in the same configuration on Precise. Thinking it might have been a regression in newer dnsmasq, I downgraded on Quantal but the problem persists.
Going back to a working Precise installation, I upgraded to the current quantal kernel from ppa:ubuntu-
Similarly on Quantal, I was able to get a functioning DHCP handshake using the Ubuntu 3.2.0-24.37 kernel, but as soon as I booted to the next published Quantal kernel (3.4.0-1.2), things were busted once again.
So it would appear something's broken since 3.3. I'm going to begin a bisect from Ubuntu-
tags: |
added: performing-bisect removed: needs-bisect |
Changed in linux (Ubuntu): | |
assignee: | Joseph Salisbury (jsalisbury) → nobody |
status: | In Progress → Triaged |
tags: | added: rls-q-incoming |
This bug is missing log files that will aid in diagnosing the problem. From a terminal window please run:
apport-collect 1035172
and then change the status of the bug to 'Confirmed'.
If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.
This change has been made by an automated script, maintained by the Ubuntu Kernel Team.