Hacker News new | ask | show | jobs
by Shicholas 2201 days ago
As someone whose gone the opposite way (moving from ECS to Kubernetes), I think the author is understating how good managed Kubernetes solutions are.

At my current job, I use Azure's managed Kubernetes service, which does a great job at providing a consistent environment that's very easily managed, no unexpected updates, great dataviz, and if you choose, simple integrations to their data storage solutions (run stateless K8 clusters if you can) and key vault. We don't do much outside of our kubectl YAML files, which as commented below has a de-facto understanding by a large number of people.

CVEs will always exist, which is why network security is important. I think we can agree that the only ingress into your cloud environments should be through API servers your team builds, and everything else should be locked down to be as strict as possible (e.g. VPNs and SSO). With a system like K8, so many eyes on the code mean so many more CVEs will exist, so I don't find this argument compelling.

My team, and so many other teams worldwide are betting that the K8 community will accelerate much faster than roll-your-own solutions, and K8 gives us the best opportunity to create cloud-agnostic architecture. Additionally, helm charts are easy to install, and afaict more software vendors are providing "official" versions - which means for a team like mine, which is happy to pay for services to manage state, in the same vain a company chooses AWS RDS over managing their own Postgres server, we can get the same benefits as the author with a cloud-agnostic solution.

1 comments

You don't see random network errors, often visible with DNS, on your Azure managed kubernetes clusters?
I haven't yet our ingresses have passed routine ping checks (we use New Relic synthetics for this) for a while now. Fingers crossed.
Ingress on AKS is easy, the outgress will be the pain if you need anything from it.