On 02/01/2013 01:37 PM, Paul Michali wrote:
No problem. So, I have some questions that I'm hoping someone can direct me as to the best place to post and how to proceed process wise on bug1059923. This is my first commit, so I'm still figuring out the process...

It takes a while. I am still trying to figure things out :)

1) The commit that was done, was reverted. Do I continue to use the review (19649) for this issue, or do I open a new review? If the latter, is there some indication in the original review that this was reverted?

You need to open a new review

2) Regarding a proposed solution, where is the best place to discuss proposals? In the review? In openstack-dev? elsewhere?

The best place is on the openstack development list (OpenStack Development Mailing List <[email protected]>). Set the subject with a prefix: [quantum]

3) Specifically on this bug… I'm unsure as to how we want to approach this (for quantum server)… whether to mark some plugin config options as 'required' so that existing logic will catch it if they are missing (I have numerous questions if that route is taken), or drive the validation logic down into the individual plugin, having them validate configuration settings (again a bunch of questions). Where can I discuss these ideas, rather than going back an forth a dozen times in the review?

The ideas are best discussed on the list.

The packing that we have done with Fedora and RHEL provides the user with a setup utility script that will make sure that the core_plugin and the database are set correctly.

I am not sure what the ubuntu guys have done (I have yet to try that out)



Thanks,

PCM (Paul Michali)


On Jan 31, 2013, at 10:23 PM, Dan Wendlandt wrote:

Apologies to Paul. quantum-core requires moderation of non-member posts (can't seem to find a way to change this), and the launchpad email notifying me of his posts were mixed in with a deluge of launchpad emails from mark's hard work tagging each quantum bug :)

sorry for the confusion.

dan

On Mon, Jan 28, 2013 at 5:42 AM, Paul Michali <[email protected] <mailto:[email protected]>> wrote:

    Any suggestions on how I tell if the sql_connection (or if there
    are other settings) are valid though?  Would there be one for the
    service itself, and maybe we could check to see if the two are at
    least the same connection type?

    Regards,

    PCM (Paul Michali)

    On Jan 28, 2013, at 8:29 AM, Gary Kotton wrote:

    On 01/28/2013 03:25 PM, Paul Michali wrote:
    OK, so where do we go from here? Do I continue to revise my
    change set to A) show plugin agent configs loaded for debugging
    purposes (like the service does), B) validate plugin agent
    config options (not specified in the bug, but people were
    suggesting this be done), and/or C) validate plugin config for
    the service?

    I am leaning towards C. Do the core_plugin validation on the service

    For B) I had raised the question in the review of how do we
    tell if the plugin specific configurations are loaded?  I see
    that many have sql_connection under DATABASE, and there are OVS
    configurations for the four agents, but all of these seem to
    provide a default value, if none specified. How do I tell if
    the default value is "valid"? I guess in the case of the bug
    they were using a MySQL database connection and without the
    plugin agent .ini it was falling back to sqlite?

    For C) similar to B) how do I pick configuration values to test
    and how do I make sure they are "valid"?

    I think that most or maybe all of the open source plugins make
    use of the DATABASE variables. A recent change was done to
    ensure that there are defined in 1 place instead of N (where
    N==number of plugins).





    Regards,

    PCM (Paul Michali)

    On Jan 28, 2013, at 6:57 AM, Gary Kotton wrote:

    On 01/28/2013 01:55 PM, Paul Michali wrote:
    On Jan 28, 2013, at 2:20 AM, Gary Kotton wrote:

    Hi,
    It has come to my attention (and not sure how I missed it),
    that the aforementioned patch does not actually solve the
    issue. There are two problems here:
    1. The validation should be done on the quantum service

    2. The agents do not require the quantum core plugin - this
    is only for the service.
    PCM: It ended up being pretty convoluted, as we went round
    and round on solutions…

    I thought that the agents did not operate correctly, if the
    core_plugin config item was not specified? I'll try to double
    check.

    The agent does not require the plugin variable. The plugin is
    actually only loaded by the service.


    For the server, it already displays the config settings
    applied, and checks that the core_plugin is specified. I
    could not identify an agent plugin config to check, as there
    are default values applied (e.g. sql_connection has SQL lite)
    and I couldn't tell if the default was valid or not.

    In any case, let me know what I need to do on this bug and on
    the commit.

    Dan has done a rollback.
    Thanks
    Gary


    Regards,

    PCM


    Sorry for the mess. How can we revert the patch?
    Thanks
    Gary





    --
    Mailing list: https://launchpad.net/~quantum-core
    <https://launchpad.net/%7Equantum-core>
    Post to     : [email protected]
    <mailto:[email protected]>
    Unsubscribe : https://launchpad.net/~quantum-core
    <https://launchpad.net/%7Equantum-core>
    More help   : https://help.launchpad.net/ListHelp




--
~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dan Wendlandt
Nicira, Inc: www.nicira.com <http://www.nicira.com/>
twitter: danwendlandt
~~~~~~~~~~~~~~~~~~~~~~~~~~~


-- 
Mailing list: https://launchpad.net/~quantum-core
Post to     : [email protected]
Unsubscribe : https://launchpad.net/~quantum-core
More help   : https://help.launchpad.net/ListHelp

Reply via email to