Twitter | Search | |
Tim Hockin
Principal SW Engineer, Kubernetes, GKE, & Google Cloud
2,236
Tweets
431
Following
10,947
Followers
Tweets
Tim Hockin 47m
Replying to @thockin
Super-congrats and thanks to the release team. I can't express enough how proud I am to work with you all, regardless of where we actually work :)
Reply Retweet Like
Tim Hockin 2h
That's 15 releases of this thing. w00t!
Reply Retweet Like
Tim Hockin Mar 22
Replying to @cmcluck @jbeda and 3 others
When a certain VP realized what we had done, he called me a not-so-nice word (with a smile :).
Reply Retweet Like
Tim Hockin Mar 17
I am thankful for the people who embrace the meta, but I also feel the need to remind people that nobody should feel forced to do that, if they prefer the eng side.
Reply Retweet Like
Tim Hockin Mar 17
...and (IMO) far less satisfying. I can slam code out 16 hours straight, but communication and reviews and planning are exhausting.
Reply Retweet Like
Tim Hockin Mar 17
I had force myself to pick up other hobbies and interests. It is something I do consciously - habits like this are pretty hard to break.
Reply Retweet Like
Tim Hockin Mar 15
This project, in particular, makes it too easy to get to this feeling. Something I have had to struggle with is simply letting go of it when I am not "on the clock". Ping me on slack, but if it is Saturday I might defer a response until Monday. And I DO NOT feel guilty.
Reply Retweet Like
Tim Hockin Mar 14
Replying to @vllry
For many meetings, at least at Google, we have a room full of people. Sig-net usually has 5-8 folks in a room, and while not exactly balanced there are a few women on the team.
Reply Retweet Like
Tim Hockin Mar 14
Replying to @vllry
At least with sig-net we have more in the room, they just don't speak so much :(
Reply Retweet Like
Tim Hockin Mar 14
Replying to @aronchick
P O U T I N E
Reply Retweet Like
Tim Hockin Mar 14
Replying to @drapenchuk @bryanl
Obviously, I am being hyperbolic, but the point stands that the API surface of the platform is larger than almost any one user needs. This is why we encourage people to build up -- a kube-supported, customized PaaS FTW.
Reply Retweet Like
Tim Hockin Mar 13
Replying to @bryanl
The analogy is apt. Even now, almost every C compiler has a way to drop to asm. The problem is finding clean ways to abstract without obliterating.
Reply Retweet Like
Tim Hockin Mar 13
Replying to @bryanl
The Kubernetes API is a data model. We have schema for things like Pods, which includes 8357 fields. If your problem is that 8357 is too many, (you want just the 10 you use), welcome to PaaS-land, it's great. If your problem is YAML/JSON, yes we will always need better tools.
Reply Retweet Like
Tim Hockin Mar 11
Is this a measured problem or hypothetical?
Reply Retweet Like
Tim Hockin Mar 11
It doesn't prevent allocation of the NodePort on nodes without backend pods, though. I think part of the question was economizing on ports.
Reply Retweet Like
Tim Hockin Mar 11
Oh, I see what you're going for. Yeah, that could have similar effects. Clever point.
Reply Retweet Like
Tim Hockin Mar 11
No, that governs the second step - where traffic goes AFTER it arrives at a node. IIUC this question is about which nodes receive traffic in the first place.
Reply Retweet Like
Tim Hockin Mar 11
Ingress is a great example of a higher level LB often built on NodePorts :)
Reply Retweet Like
Tim Hockin Mar 11
Short answer, no not really. NodePorts were not really designed for end users, but for providers to build higher order LB functions.
Reply Retweet Like
Tim Hockin Mar 8
Replying to @mattomata
I was going to say that "Matt"s are a dime a dozen, but to get both first and last names is pretty sweet. :)
Reply Retweet Like