I just noticed this review:


And gave it some real thought. This will likely break any large scale
usage of signals, and I think breaks the user expectations. Nobody expects
to get a failure for a signal. It is one of those things that you fire and
forget. "I'm done, deal with it." If we start returning errors, or 409's
or 503's, I don't think users are writing their in-instance initialization
tooling to retry. I think we need to accept it and reliably deliver it.

Does anybody have any good ideas for how to go forward with this? I'd
much rather borrow a solution from some other project than try to invent
something for Heat.

I've added Marconi as I suspect there has already been some thought put
into how a user-facing set of tools would send messages.

OpenStack-dev mailing list

Reply via email to