On 07/08/2012 10:48 AM, Salvatore Orlando wrote:
At a first glance (meaning that I've not looked in detail at the code), I would say that this kind of situations will be much more likely to happen once we start attaching different services to Quantum ports. I reckon this is an example of incompatibility between the dnsmasq-based plugin/driver for DHCP and the linux bridge L2 plugin - there will be of course cases in which plugin are definitively incompatible, but we should strive to make sure that such incompatibility cases are limited - and port naming should probably not be a reason for incompatibility.

Finding a final solution during the Folsom timeline is going to be quite difficult, and it certainly needs a thorough discussion. Among the options suggested by Gary, I probably prefer (in order) either #4 or #2. From my (limited) perspective, it looks like it could be merged as a separate bug fix, do you agree?

I do not think that this is a blocking issue and can certainly be handled as a bug fix. I have given +1 to the development (+2 is pending a really minor issue).

I would prefer that #4 be the solution. The reason is twofold:
1. I do not think that a agent should create a device from the network id. I think that this is a bug in the linux bridge implementation and that it should be removed as part of the V1 removal. 2. As a rule I do not think that the addition of new agents should require us to change the functionality of existing agents. This is not written in stone but should be taken into account.


Salvatore

On 8 July 2012 08:34, Gary Kotton <gkot...@redhat.com <mailto:gkot...@redhat.com>> wrote:

    Hi,
    Great work regarding the agent. My tests with the OVS have passed
    successfully. I nonetheless have some problems regarding the
    linuxbridge. The problem is that if a device is created from the
    network ID then it should have a "gw-" as the prefix.
    I think that if you create the dnsmasq interface from the port id
    then this will solve the issue. This would require a few changes
    to the dhcp code. If not then we need to change the logic of the
    linux bridge agent. I have added my comments to
    https://review.openstack.org/#/c/9064/.

    Off the bat I think that there are a few options:
    1. DHCP agent changes tap device to "gw-..." no changes to linux
    bridge. The fact that the DNSMASQ may not be the default gateway
    may be a bit misleading (I agree with what you have written in
    your comments)
    2. Change linux bridge code to check if the tap device is contains
    a network id.
    3. Add a new prefix that could indicate that the device is a
    "dhcp" device. This is a minor change in the linux bridge agent to
    support. I'd be happy to do this.
    4. DHCP agent uses the port id. No changes to linux bridge.
    Any thoughts?

    Thanks
    Gary

-- Mailing list: https://launchpad.net/~netstack
    <https://launchpad.net/%7Enetstack>
    Post to     : netstack@lists.launchpad.net
    <mailto:netstack@lists.launchpad.net>
    Unsubscribe : https://launchpad.net/~netstack
    <https://launchpad.net/%7Enetstack>
    More help   : https://help.launchpad.net/ListHelp



-- 
Mailing list: https://launchpad.net/~netstack
Post to     : netstack@lists.launchpad.net
Unsubscribe : https://launchpad.net/~netstack
More help   : https://help.launchpad.net/ListHelp

Reply via email to