3 ms·

I felt like if anything went wrong I'd be in a world of hurt trying to debug it. This is the part very few k8s aficionados mention: how complicated is it to tr

I felt like if anything went wrong I'd be in a world of hurt trying to debug it. This is the part very few k8s aficionados mention: how complicated is it to troubleshoot when something goes wrong?

geggam 8y ago
Not to mention when folks talk HA and you ask about federation they just give you a blank stare. k8s is HA in the same region if you are using GKE. Not really what you can call HA
- pests 8y ago
  Google just posted a blog[0] about multi-region clusters and ingress. It's a little hacky as you need to create multiple clusters in each zone/region and then with an anycast IP attached to a Google Cloud Load Balancer. [0] https://cloudplatform.googleblog.com/2018/06/How-to-deploy-geographically-distributed-services-on-Kubernetes-Engine-with-kubemci.html https://cloudplatform.googleblog.com/2018/06/How-to-deploy-g...
  - geggam 8y ago
    Hacky is not production ready.
    - pests 8y ago
      Its not hacky in that sense. They provide a new kubemci tool and I'm sure it will be given the k8 release treatment.
barrkel 8y ago
The core k8s, not that bad, once you've learned where to look - logs on pods, events, and kubelet logs from journalctl. That last one is important not to forget when debugging plugin controllers.