Hacker News Reader

Comment by ratherbefuddled | Hacker News Reader

I don't like the direction this is heading, it seems like the SLA and the accompanying charge could easily have been optional.

We have a couple of dozen clusters, two per client, and can't change the architecture. We use helm and terraform and can build new clusters quickly but we can't treat them entirely like cattle because we don't own all the DNS. Our clients are not the sort to do things quickly - or even slowly.

Does anybody have any good and up to date resources comparing the current options for K8 providers? I'd like to get a feel for what it would take to switch.

View on HN

As many have mentioned here already, $72/mo most likely a rounding error on workloads kubernetes is designed for.

I think, most customers love the change because of SLA where even 1 minute of downtime per year is amplitude more costly than 10 years of cluster managing fee.

This also showing the commitment from google to provide great and reliable service.

If you're looking to run k8s "for free", Digital Ocean looks like the way to go, but again, it's two completely different set of offerings and if you've chosen the google cloud in the first place then DO doesn't look like suitable alternative.

> As many have mentioned here already, $72/mo most likely a rounding error on workloads kubernetes is designed for.

There are _many_ reasons to use k8 beyond just workload scale and that amount per cluster per month isn't anywhere near a rounding error for many deployments.

I would be interested to know when you may need kubernetes for small types of workload and simple architecture.

For my understating, support contract alone for google cloud will cost you around $150/mo.

For small to medium workloads there are plenty of tools if you want to use containers: docker swarm and nomad. Docker Swarm is really simple and most engineers already know it because of `docker-compose.yml` they're using each day.

I can't really understand what type of workload you have because kubernetes cluster management requires you to have at least 2 full-time DevOps team. What is $72 compared to salary of two employees?

If you have an issue with $72 for SLA guarantee then I can't really understand why you need google cloud at all. They are literally bound to maintain zero downtime because of huge losses if something go down for their entire customer base.

One simple example - of many: Enterprise clients with low workload but very sensitive data strict infosec and segregation requirements - several clusters per client all managed with infrastructure as code.

There are lots of benefits to k8 other than just scale, and there are architecture choices that are made for good reasons that rely on separate clusters with no particular need for a legal uptime SLA on a control plane.

This decision will cost us a lot more than $72 a month.

Reply on news.ycombinator.com