Apiman 4 — what do you want to see?

[msavy]

Hi. What features and changes would you like to see added to Apiman 4?

[OSMOSA44]

Hi Marc,

From my point of view,

The next generation of gateway should be as autonomous and immutable as possible to facilitate its integration in a clustered environment such as kubernetes,

Some ideas below,

• [Gateway] Evicting Elasticsearch by working on a full headless gateway
• [Manager] Tool to export organization configuration (json) to integrate in headless gateway
• [CI/CD] for each new configuration, construct new docker image with only needed plugins policies and client certificate (mTLS)
• [Gateway] Send metrics to stdout and catch them with a monitoring tool such as datadog through docker logs integration for example
• [R&D] Build a rate/quota limit service that works through multiple kubernetes clusters in multiple datacenters
• ...

Regards,
Laurent

[msavy]

Happy New year, Laurent! I would be delighted to have a discussion with you and your organisation about these requirements/roadmap items and more 👍. I'm sure we'll do that soon. Replies inline:

[Gateway] Evicting Elasticsearch by working on a full headless gateway

This is already possible via the headless gateway registry, but perhaps we could make it more convenient.

There are a few cool patterns I've seen Apiman users using to provide a more immutable gitops-style deployment. I'll probably blog on this when I have time, or perhaps provide a set of skeletons/examples via BPL Enterprise.

The great thing about this style is that it gives you best of both worlds; by default everything is a clone that you deploy from config, but you can log in and make manual changes in the "traditional" way if you want for testing, monitoring, or other reasons.

[Gateway] Send metrics to stdout and catch them with a monitoring tool such as datadog through docker logs integration for example

In terms of logging out to file (or stdout if you want), this will be possible in 3.1.0.Final.

I've been considering offering specific Datadog metrics integration through my company (Black Parrot Labs). Perhaps that would be interesting?

[R&D] Build a rate/quota limit service that works through multiple kubernetes clusters in multiple datacenters

Sounds like a very interesting requirement. We should discuss this in detail! I've been working on an improved rate limiting implementation that is specifically designed to operate in a distributed, high-load environment without hammering the database and avoiding most latency/throughput tradeoffs (via Black Parrot Labs Enterprise service, currently).

It would likely be valuable to have a detailed discussion on this topic to understand your requirements, as multi-datacentre is more complicated in terms of engineering tradeoffs and figuring out where we can be flexible.

[msavy]

And while I remember:

I've been working on some detailed plans to further relegate Elasticsearch to more metrics-specific workloads, but still allow the data acquisition that people value (e.g. knowing what the current state of a rate limit is, etc).

There's already the probes functionality that I added in Apiman 3.x as a preview, but I believe I can extend this further with an event-driven approach that is more amenable to a push-style monitoring solution.

More on this in future and/or during our discussions!

[msavy]

I am continuing to work hard on Apiman 4...

Thoughts and input continue to be welcome. I'm interested in continuing to develop relationships with partners, too.