greptilian logo

IRC log for #sourcefu, 2018-03-12

http://sourcefu.com

| Channels | #sourcefu index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

All times shown according to UTC.

Time S Nick Message
14:53 dotplus at work, we're killing rancher-based kubernetes in favour of kubernetes
14:55 pdurbin Huh. Why?
14:57 dotplus mostly because rancher is extra complexity for little extra benefit. Both our infra and our apps _must_ be "As Code", so the webui(s) are only going to be useful for visualization - they are ~readonly.
14:58 dotplus remember rancher started as a sort of abstraction layer between various container scheduling approaches. well, now that we've picked k8s instead of swarm/mesos/whatever, we no longer need that abstraction/portability
14:58 semiosis joined #sourcefu
15:00 pdurbin Right. On Friday I learned that Rancher is older than Kubernetes. I added this to some notes I took at https://docs.google.com/document/d/12njDDHHfoNdos0By3lKlxYbRWLsAp8Eo-AGpjuBF_rY/edit?usp=sharing
15:27 semiosis joined #sourcefu
18:16 bear we were thinking of using rancher until we had the same thoughts - with infra-as-code, using helm+k8s was cleaner
18:16 bear we are also trying to remove GUI's as change agents
19:19 dotplus nice - I'm stealing that phrase.
19:30 dotplus we're not currently using helm, just writing (generating/templating) k8s manifests directly, although I believe someone in my org is looking into it. Does helm provide a powerful&useful abstraction layer/DSL around the k8s API? I haven't looked into it because when I first heard of it, it was made out [to me] to be just syntactic sugar for newbies so I kinda ignored it. Is that a mistake?
19:42 bear it does hide a lot of the gnarly bits - but it doesn't fully remove the need for knowledge of k8s
19:42 bear it's best when given as template examples to devs by SRE folk - we use it to mitigate mistakes in k8s templates that cause balancing issues
19:43 bear it also allows us to give an air gap between the code deploy and the k8s push
19:43 bear the code deploy generates the helm chart, that is stored and then a message triggers it's actual application
19:44 bear which also allows our ops folk to be able to roll back without having to push a commit
20:06 pdurbin I heard a bit about Helm at https://thewomenintechshow.com/2018/02/26/learning-kubernetes-with-amy-chen/
20:07 pdurbin "learned to use Kubernetes by building helm charts"
20:08 pdurbin Lots of apps listed at https://github.com/kubernetes/charts

| Channels | #sourcefu index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

http://sourcefu.com