Hacker News Reader

Comment by jedberg | Hacker News Reader

If you're interested in dirt simple Lambda deployment, check out Kappa[0], written by the same person as boto.

Gordon looks interesting and I'm definitely going to check it out, but from what I can tell, one of the downsides is that it uses CF.

Don't get me wrong, I love CF for it's reproducibility, but waiting for a CF stack to run on every deploy will be frustrating for rapid iteration. CF is great for managing your infrastructure once you have a good idea of how it should look and aren't making constant changes anymore.

I suspect my workflow will be something like using Kappa while I'm iterating, and then using Gordon for the final deployment once it looks like everything is working as expected.

[0] https://github.com/garnaat/kappa

View on HN

Can Kappa support custom versions and aliases? That's been one thing I haven't seen able to be controlled via kappa config or command line, and it's been the reason I haven't used it.

It doesn't support custom versions and aliases yet. Curious what your use case is.

It sort of does custom aliases by giving the function an alias with the environment name, but that probably isn't what you want.

I've got multiple lambda functions that need to be associated with each other for an AWS Simple Work Flow. For CI I've it set up to update the versions of each lambda in a project repo. During packaging and deployment I inject the configuration of the git sha so that the next lambda function in the workflow that is called is consistent. Once the new workflow lambda call-chain is in place I can test / validate the new environment and then roll over the lambda that starts the workflow.

What I need to make sure of is that all the lambda functions / roles are updated before the starter so that the workflow doesn't break, as there could be breaking changes between commits.

A suggestion would be to create a repo per lambda function, but that becomes a hassle to manage after a few dozen, and just throws in even more work on ensuring the order of updates necessary. It is easier to flow between commit versions.

Open an issue as to how that would work and I can take a look (I have commit bit on Kappa now).

I've only used serverless and that only needs to run CF when you update backend resources.

Yeah serverless is a little smarter and directly uploads lambda functions.

The main reason I don't use serverless is because it's geared towards web development -- the lambdas and the api gateway are tied together. Most of what I'm doing is not web based.

Reply on news.ycombinator.com