Twitter | Search | |
Brian Grant
GKE co-TL, Kubernetes Steering Committee member, SIG Architecture co-Chair, CNCF TOC member
638
Tweets
213
Following
4,518
Followers
Tweets
Brian Grant Aug 16
Kapps, ArgoCD, AppController, Spinnaker, and other deployment tools implement their own resource-type-specific progress, status, and wait logic. It's very error-prone and not-extension-friendly, so that's something that needs work.
Reply Retweet Like
Brian Grant Aug 16
Right now, cli-utils is for exercising next-gen kubectl and apiserver functionality like server-side apply () and a safer, more standard model for apply --prune, which provides declarative deletion ()
Reply Retweet Like
Brian Grant Aug 16
I agree (see ), but first we're implementing it on the client side:
Reply Retweet Like
Brian Grant Aug 16
Replying to @dbentley @googlecloud
Reply Retweet Like
Brian Grant Aug 16
Speaking of configuration, you can now use Kubernetes CRDs to manage some resource types, with more to come. This was shown at Google Next, but is now beta, so give it a try and let us know what you think:
Reply Retweet Like
Brian Grant Aug 10
Replying to @dims @jerenkrantz and 6 others
Often the interaction is through github. I didn't see a comment from you on the closed issue you cited. Did you file a new issue? It looks like filed a new one:
Reply Retweet Like
Brian Grant Aug 9
Agree that we still don't have nearly enough diagrams
Reply Retweet Like
Brian Grant retweeted
Brad Topol Aug 6
It’s the time of year when the SIG Docs team would like your input on Kube documentation pain points that you would like to see addressed. If you know of some good pain points let me know! (Plz RT)
Reply Retweet Like
Brian Grant Aug 1
v2 isn't necessary in this case. PodDisruptionBudget isn't v1 yet.
Reply Retweet Like
Brian Grant Aug 1
Yeah, I freeze all issues I care about. I froze that one.
Reply Retweet Like
Brian Grant Aug 1
Reply Retweet Like
Brian Grant Aug 1
Thanks. My suggestions:
Reply Retweet Like
Brian Grant Aug 1
What is the issue number? K8s has tons of room for improvement
Reply Retweet Like
Brian Grant Aug 1
This is a feature request, not a problem with the overall design of Kubernetes, pods, or controllers. PDB is still immature. File an issue if you don't find one, and comment on . The decoupling has been a huge win.
Reply Retweet Like
Brian Grant Jul 24
I will try to Twitterize some of the config docs, since that is a popular topic
Reply Retweet Like
Brian Grant Jul 24
Replying to @steren @GCPcloud and 3 others
So many docs, so little time... These days a lot of what I do is play matchmaker: team X, you should talk to team Y.
Reply Retweet Like
Brian Grant Jul 22
Replying to @sszuecs
Unified cgroup hierarchy forces some changes also
Reply Retweet Like
Brian Grant Jul 22
Replying to @sszuecs
Other than latency tolerance, Borg was even more implicit, and had too many overlapping control loops. I think the K8s model is a better starting point but has had less iteration, because it's 10 years younger, and also not the same priority when machines can also be scaled
Reply Retweet Like
Brian Grant Jul 22
Replying to @bgrant0607
Ok, this topic doesn't fit into a Twitter form factor very well. Maybe some day I'll get around to writing this up more in long form. For now, that's about all I have time for, but questions are welcome
Reply Retweet Like
Brian Grant Jul 22
Replying to @bgrant0607
If managing cluster-level sharing using ResourceQuota and LimitRange, oversubscription can be done at that level also. The original designs were described by and , with improvements in
Reply Retweet Like