[vcap-dev] NATS protocol

Perhaps we can coerce Luke Bakken, who studied these calls closely to build an NG compatible Windows DEA for Tier 3, to write a few notes on that?

(Coughs)

Sent from my iPhone

There was a web app on cf.com (api-doc.cf.com but its dead now) that discussed the NATS APIs for v1. Perhaps a CF admin could find and restart that app?

Source is https://github.com/paasio/apidocs

Hey everyone -

I had the same thought as Matt when I read Gandalf's message this morning.
I'll see what I can document.

Off the top of my head -

Gandalf - the messages are very similar to the v1 CC - DEA with the
addition of the staging.* messages to stage apps using buildpacks.

I use the nats-sub binary from https://github.com/derekcollison/nats frequently
to subscribe to specific messages and to watch all messages.

For instance, this will run the program from source and will subscribe to
all messages:

$ cd ~/src/nats/
$ bundle install
$ cd bin
$ bundle exec nats-sub '>' -s nats://10.0.0.1:4222

Or, you can change the subscription to dea.* to just see DEA-directed
messages.

Luke

Reading the source and the specs and watching on the bus are the best ways
to figure it out. The format of the messages isn't set in stone, it
occasionally has to change to accomodate new features.

Also, nats-sub is good but [nats-sub-pp](https://github.com/mmb/nats_sub_pp)
is better.

2013/6/2 Matthew Kocher <mkocher@pivotallabs.com>
I'll try to read the source but what I would like to know is how the caller
will be notified about the receiver response.
For example, if CCNG ask DEAs for a deploy through NATS, how CCNG will be
able to get the DEA response? Are there any "response" event triggered by
DEA and subscribed by CCNG ?

My old API site is still available here: http://apidocs.cloudfoundry.com/
(no dash)

It hasn't been updated in a little over a year though... on v1, some things
had changed with starting an app to move away from the help request
followed by start call. And I don't know if the messages have changed with
V2... my guess is it includes more metadata now. It doesn't look like I'd
gotten to documenting the DEA though.

I used to simply run "nats sub '>'" to get anything and everything.
  Sometimes I'd filter stuff with a "grep -v" for component
discover/announce messages.

From how things used to work, the CC was never concerned with whether or
not the app actually started. That was the responsibility of the Health
Manager. When the start process finished (either success or failure), the
DEA would send a heartbeat containing just that app to the HM. It would
look like
this: http://apidocs.cloudfoundry.com/health-manager/subscribe-dea-heartbeat

Ken

Thanks Ken.


​It would be awesome for the people who are investigating the internal messaging/APIs to updates Ken's docs repo!