Today every self-respecting developer thinks of microservices and containers. This approach significantly simplifies deployments and maintenance efforts, while shortening the development-to-production cycle length.
But there are some quite weakly addressed pain points on that route: we can reliably deploy a product as a swarm of weakly-linked containers, but we're lacking most of the usual tools to get insights of what's happening inside those containers|services, and how to debug and troubleshoot our production issues.
As a team in the middle of such transition, we at SCC have hit most of those issues:
- Any incoming request requires several containers to work in unison. How to log that work so it's easy to retrieve everything what happened on all the affected nodes?
- Sometimes we see a drop in our API performance. Is it really a drop or a glitch? What's its cause? Where's the bottleneck?
- Where should we look to make sure everything is green in our cluster? What dashboard should we use, what healthchecks should be there, how to send notifications if something goes wrong?
That's an interesting topic, the containers team can provide assistance with that too