Datasource Fallbacks (simplestreams)

[Andrew Wilkins]

I have landed the change. If this causes problems, we can add a flag
private-cloud datasources to prevent fallback.

We'll need to document this in the release notes. I have created a new bug
so this gets a mention in 1.17.0:
https://bugs.launchpad.net/juju-core/+bug/1247730

Cheers,
Andrew


On Thu, Oct 10, 2013 at 4:34 PM, John Meinel <john at arbash-meinel.com> wrote:

> So there is a difference between there not being a simplestream source at
> a location and there not being a matching entry in an existing file. If
> someone sets up a private stream falling back might not be expected.
>
> John
> =:->
> On Oct 10, 2013 11:16 AM, "Andrew Wilkins" <andrew.wilkins at canonical.com>
> wrote:
>
>> On Fri, Oct 4, 2013 at 9:24 PM, Curtis Hovey-Canonical <
>> curtis at canonical.com> wrote:
>>
>>> Do we place a specific warning in the log. If I bootstrap with
>>> --show-log, do I see a message that clearly states the image cannot be
>>> found and and that Juju is falling back? I want to know if I have
>>> misconfigured something.
>>>
>>
>> Not a warning, because it's not necessarily unexpected. The code
>> currently logs a DEBUG level message if it can't find product (image,
>> tools) information from a source.
>>
>> Cheers,
>> Andrew
>>
>>
>>>
>>> On Fri, Oct 4, 2013 at 6:24 AM, William Reade
>>> <william.reade at canonical.com> wrote:
>>> > Yeah, let's go with always-fall-back. Much more likely to be useful in
>>> > general, and we can worry about masking off the rest of the world if
>>> there's
>>> > a clear need for it that can't be easily addressed outside juju (if the
>>> > environment can't see a stream, it won't be able to fall back to it
>>> ;)).
>>> >
>>> >
>>> > On Fri, Oct 4, 2013 at 11:26 AM, Jeroen Vermeulen
>>> > <jeroen.vermeulen at canonical.com> wrote:
>>> >>
>>> >> On 04/10/13 12:26, John Arbash Meinel wrote:
>>> >>
>>> >>> We came to this because the Azure provider was adding the "daily"
>>> >>> stream to the search path. Which meant that we would find the daily
>>> >>> stream, parse it, not find anything and then stop. Without falling
>>> >>> back to the releases stream which has the actual entries we are
>>> >>> looking for. (We "solved" this by just removing the daily stream,
>>> >>> which was also for other reasons.)
>>> >>
>>> >>
>>> >> I did that, IIRC at a time when only the daily images worked yet but
>>> >> releases were the obvious long-term way to go.
>>> >>
>>> >> The implied contract for the imagemetadata package seemed to be that
>>> it
>>> >> would consider only the selected stream, and ignore any other streams
>>> it
>>> >> found in the search path.  From what you say, that's what Andrew
>>> >> implemented.  Otherwise, what's the point of supporting a search path
>>> >> consisting of multiple URLs?
>>> >>
>>> >> With the fallback, having both the daily and the release stream in the
>>> >> path allows users to select one or the other by setting just a stream
>>> name —
>>> >> no messing with simplestreams URLs in the most common case.
>>> >>
>>> >>
>>> >>
>>> >> Jeroen
>>> >>
>>> >>
>>> >> --
>>> >> Juju-dev mailing list
>>> >> Juju-dev at lists.ubuntu.com
>>> >> Modify settings or unsubscribe at:
>>> >> https://lists.ubuntu.com/mailman/listinfo/juju-dev
>>> >
>>> >
>>> >
>>> > --
>>> > Juju-dev mailing list
>>> > Juju-dev at lists.ubuntu.com
>>> > Modify settings or unsubscribe at:
>>> > https://lists.ubuntu.com/mailman/listinfo/juju-dev
>>> >
>>>
>>>
>>>
>>> --
>>> Curtis Hovey
>>> Canonical Cloud Development and Operations
>>> http://launchpad.net/~sinzui
>>>
>>> --
>>> Juju-dev mailing list
>>> Juju-dev at lists.ubuntu.com
>>> Modify settings or unsubscribe at:
>>> https://lists.ubuntu.com/mailman/listinfo/juju-dev
>>>
>>
>>
>> --
>> Juju-dev mailing list
>> Juju-dev at lists.ubuntu.com
>> Modify settings or unsubscribe at:
>> https://lists.ubuntu.com/mailman/listinfo/juju-dev
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju-dev/attachments/20131105/6fa1c764/attachment-0001.html>