As an RDS broker user, I want to get alerts around my database

[afeld]

We should give users an option to (and/or enable by default?) alerts around high levels or spikes in:

• Memory
• CPU

@LindsayYoung Anything I'm missing?

[LindsayYoung]

free space

[mogul]

The current quarter's direction is to give you the ability to get the metrics out, but then leave who/how to alert up to you (where OpenDuty runs easily on Heroku, and is therefore highly likely to also run on cloud.gov...)

[jmcarp]

I took another look at this, and I think it's going to be more work than expected to expose metrics. Cloudwatch metrics don't expose resource-level permissions, so we can't make iam policies that grant the user permissions to get metrics from a particular rds instance--we'd have to run a web service that proxies the aws api with a custom auth layer. We could also set the broker up to allow users to configure arbitrary alerts via sns, but that still wouldn't expose the actual metric values.

Thoughts @mogul? Copying @cnelson because we might have a similar issue with cloudfront and federalist.

[mogul]

Not doing this any time soon.