Asynchronous deployments?

[Martin Meredith]

My title probably doesn't make sense. So I'll try and explain. As far as I can see, heaven doesn't allow for a workflow similar to the following:-

• Receive deploy event from Github
• Trigger deploy process.
• Wait for response (pingback) to say this has been done/failed
• Send Deploy Status Event to Github

As far as I can tell at the moment, when the deploy is triggered, resque runs, and is expected to return a pass/fail which it immediately returns to Github.

I'd want to be able to do something like, create an AMI in AWS. This would take a significant amount of time, and I can have SNS send a notification (or similar) when this (long running) process is done. However, as far as I can see, Heaven won't cater for this, and has to sit around polling for the AMI build to be finished.

Am I missing something here? If I'm not - can you give me any pointers for the best way to go about adding this in?

[Corey Donohoe]

Your understanding is correct. Right now you'd have to poll waiting to see if it was completed. If you want to write something like this you'd need to.

• Write a provider that fires off the request to AWS or whatever.
• Fire off a deployment status event to github that leaves it in a pending state.
• Add a route and controller action to handle some kind of callback url for it.
• Handle the response and save the deployment status as success/failure.

Unless we can decide on the pingback/callback url being somewhat generic it probably wouldn't be a good fit for adding to the project. Stuff like that would be good for a custom fork or something similar.

[Martin Meredith]

Yes, I was thinking of something generic for the callback, so that multiple
asynchronous callbacks are available (I want to do multiple things like
this).

I'll have a play and see what I come up with, hopefully then you can
suggestions for what is useful for the project, and what is useful for a
custom fork.
On 8 Jul 2014 18:11, "Corey Donohoe" [e-mail] wrote:

Your understanding is correct. Right now you'd have to poll waiting to see
if it was completed. If you want to write something like this you'd need to.

• Write a provider that fires off the request to AWS or whatever.
• Fire off a deployment status event to github that leaves it in a
  pending state.
• Add a route and controller action to handle some kind of callback
  url for it.
• Handle the response and save the deployment status as
  success/failure.

Unless we can decide on the pingback/callback url being somewhat generic
it probably wouldn't be a good fit for adding to the project. Stuff like
that would be good for a custom fork or something similar.

—
Reply to this email directly or view it on GitHub
#50 (comment).