Open Stack

Hi Rico,

Firstly sorry for the slow reply!  I am finally catching up on my
backlog.

Rico Lin <rico.lin.guanyu at gmail.com> wrote:
>Dear all
>
>Back to Vancouver Summit, Ifat brings out the idea of integrating Heat,
>Vitrage, and Mistral to bring better self-healing scenario.
>For previous works, There already works cross Heat, Mistral, and Zaqar for
>self-healing [1].
>And there is works cross Vitrage, and Mistral [2].
>Now we plan to start working on integrating two works (as much as it
>can/should be) and to make sure the scenario works and keep it working.
>The integrated scenario flow will look something like this:
>An existing monitor detect host/network failure and send an alarm to
>Vitrage -> Vitrage deduces that the instance is down (based on the topology
>and based on Vitrage templates [2]) -> Vitrage triggers Mistral to fix the
>instance -> application is recovered
>We created an Etherpad [3] to document all discussion/feedbacks/plans (and
>will add more detail through time)
>Also, create a story in self-healing SIG to track all task.
>
>The current plans are:
>
>   - A spec for Vitrage resources in Heat [5]
>   - Create Vitrage resources in Heat
>   - Write Heat Template and Vitrage Template for this scenario
>   - A tempest task for above scenario
>   - Add periodic job for this scenario (with above task). The best place
>   to host this job (IMO) is under self-healing SIG

This is great!  It's a perfect example of the kind of cross-project
collaboration which I always hoped the SIG would host.  And I really
love the idea of Heat making it even easier to deploy Vitrage
templates automatically.

Originally I thought that this would be too hard and that the SIG
would initially need to focus on documenting how to manually deploy
self-healing configurations, but supporting automation early on is a
very nice bonus :-)  So I expect that implementing this can make lives
a lot easier for operators (and users) who need self-healing :-)

And yes, I agree that the SIG would be the best place to host this
job.

>To create a periodic job for self-healing sig means we might also need a
>place to manage those self-healing tempest test. For this scenario, I think
>it will make sense if we use heat-tempest-plugin to store that scenario
>test (since it will wrap as a Heat template) or use vitrage-tempest-plugin
>(since most of the test scenario are actually already there).

Sounds good.

>Not sure what will happen if we create a new tempest plugin for
>self-healing and no manager for it.

Sorry for my ignorance - do you mean manager objects here[0], or some
other kind of manager?

[0] https://docs.openstack.org/tempest/latest/write_tests.html#manager-objects

>We still got some uncertainty to clear during working on it, but the big
>picture looks like all will works(if we doing all well on above tasks).
>Please provide your feedback or question if you have any.
>We do needs feedbacks and reviews on patches or any works.
>If you're interested in this, please join us (we need users/ops/devs!).
>
>[1] https://github.com/openstack/heat-templates/tree/master/hot/autohealing
>[2]
>https://github.com/openstack/self-healing-sig/blob/master/specs/vitrage-mistral-integration.rst
>[3] https://etherpad.openstack.org/p/self-healing-with-vitrage-mistral-heat
>[4] https://storyboard.openstack.org/#!/story/2002684
>[5] https://review.openstack.org/#/c/578786

Thanks a lot for creating the story in Storyboard - this is really
helpful :-)

I'll try to help with reviews etc. and maybe even testing if I can
find some extra time for it over the next few months.  I can also try
to help "market" this initiative in the community by promoting
awareness and trying to get operators more involved.

Thanks again!  Excited about the direction this is heading in :-)

Adam