Who here loves yaml?Been to a couple of Linux conferences this year and almost every product seemed to have some integration with or use for kubernetes. At Velocity Conf in San Jose 4/5 keynotes we're about Kubernetes.
9/7/2019 5:34:10 PM
k8s was so 2018[Edited on September 7, 2019 at 7:25 PM. Reason : I do love yaml though. and also k8s. but docker more than k8s][Edited on September 7, 2019 at 7:26 PM. Reason : I miss having a mini-HN crowd here. Was in a nodejs thread yesterday. We don't talk about stuff here]
9/7/2019 7:23:57 PM
I work with a lot of on-prem web app deployments, and seems like a lot of people (higher ed mostly) are not really prepared to handle docker or kubernetes based apps.Is the use case mostly cloud services?
9/8/2019 12:01:44 AM
Kubernetes changes the workflow from:Developer needs to deploy a new service or Cron job which requires a database.Get techops and SysOps to provision new machines. Do capacity planning in data centers. Lay down the OS and setup monitoring and logging. Write wrapper scripts to start and stop the services. Ensure current base images have everything needed by the new service. Then tell the developers it's ready to go and setup some kind of workflow in a ci/cd pipeline to deploy the service.Instead with kubernetes all the base os work and monitoring and alerting is decoupled from the developers. So the developer writes a yaml file which effectively tells kubernetes both how many and what to run and any disks or other storage it needs. They don't have to wait on SysOps to get the machines ready.The first chapters of the Google SRE book has some good info about the basic architecture: https://landing.google.com/sre/sre-book/chapters/production-environment/In addition it creates a standard way to deploy into the cloud providers and bare metal. If you hear people talking multicloud or hybrid cloud and products to help, most are all based on K8s.[Edited on September 8, 2019 at 9:40 AM. Reason : A]
9/8/2019 9:30:53 AM
k8s is mostly a leaning tower of abstraction, with the exception being working on prem. you can't just drag and drop your stuff between sites like you'd hope. on cloud however, you have most of these abstractions already in place and infrastructure as code facilities that are better integrated with the platform services as well. the majority of compute may still be on prem, but the majority of innovation happens on cloud, so 90% of my work is on cloud.
9/8/2019 11:43:10 AM
Have a few hundred nodes going, working towards 1000. 30,000 to go. We are deploying bare metal, AWS and to own vm stack consistently. The main differences are in DNS and how disk provisioning works. Same from yaml/deployment perspective.Been deploying with gitlab and monitoring with Prometheus. Debating building a slack exporter that does counts of post by user and channel in slack. Would then alert if rate exceed x/hour. Integration with workday to fire employees who violate this condition.It will be a good April fool's joke.[Edited on February 20, 2020 at 1:07 AM. Reason : A]
2/20/2020 1:05:46 AM
lol, at my last company a friend and I had a slack:commit ratio that was a pretty good indicator of who wasn't doing any work. probably should have made more HR decisions based on it
2/20/2020 2:47:11 PM
Yeah, the GitHub exporter would be awesome as well.
2/21/2020 6:25:45 PM
is anyone using openfaas on k8s?
3/2/2020 12:56:06 AM
We did, what is your use case for it? We've since been down the rabbit hole on Knative, Rio, and most of the others. OpenFaas doesn't really offer much outside their templates, which are really just doing a docker build under the hood.
3/2/2020 9:03:59 AM
building an event driven platform that's portable to on-prem. I like the queueing, async + sync invocations, and automagic rest wrapper. at a high level is basically cgi for k8s, but I value not having to write it and just do some additions on the message bus side[Edited on March 2, 2020 at 11:41 AM. Reason : my use case is I need an orchestration layer for arbitrary binaries]
3/2/2020 11:39:45 AM