zwachtel.bsky.social
@zwachtel.bsky.social
Anyways listening to this episode brought great catharsis in both seeing others in the space seeing similar issues and listening to effective debugging to get to root cause.
April 4, 2025 at 6:18 PM
I think that customer networks are the hardest part of this space.

I was hoping that Oxide would avoid some of this pain as it’s a closed box rack appliance with the ToRs integrated but I guess the problems just peel back one hop.
April 4, 2025 at 6:15 PM
Also thanks for the pointer, knew there had to be a better way and was wondering what that k8s prefix signified.

Think I even glossed over that variable and didn’t read the comment in depth 😅.
February 16, 2025 at 1:30 AM
Unfortunately I only discovered Auger after coding this.

Moved to using Auger for debugging the type of issues I created this for.

Anyways it was a good opportunity to dig into how k8s stores resources in etcd.
February 16, 2025 at 1:14 AM
I’m sure this is touched on in 10, but one thing folks always get hung up on is when/if it is ok to ever use information stored in a status to make decisions in the reconcile loop.
January 19, 2025 at 5:32 PM
This is such a good point.

It’s a much easier exercise to come up with the inputs to an abstraction layer than to figure out the outputs.
January 16, 2025 at 4:15 AM
I have also seen my fair share of people running etcd on “unacceptable local disks”, and wondering why a strong wind was toppling over their cluster…

Luckily in 2024, the number of cases I have seen of this has dropped tremendously since experiencing my first etcd headaches ~2018.
December 18, 2024 at 4:06 PM
In his defense … 5+ years later the private cloud space is still attempting this approach…
December 3, 2024 at 3:46 PM
Hmm I live in NYC and hate DNS…
December 3, 2024 at 1:29 AM
I thought the Chicago cocktail was a beer/shot combo of Old Style and Malört…
November 24, 2024 at 12:33 AM
Or i guess “OxKS”
November 21, 2024 at 7:05 PM
Can’t wait for OKS!
November 21, 2024 at 7:05 PM
In general, I feel like kubecon had a huge disconnect between all the topics on stage and what was happening in the hallway.

My hallway track was full of tool fatigue, cluster horror stories and a sense of not knowing where to go next.
November 17, 2024 at 8:34 PM
Oh wow… may need to sprint to this from the airport!
November 12, 2024 at 2:19 AM
Unfortunately this lesson is only learned the hard way…
October 24, 2024 at 9:12 PM
Oh interesting. First time I am reading about “wide events” and I love it.

My team calls something like “wide events” telemetry and kusto is our storage/queryengine.
October 24, 2024 at 12:13 PM
Do people still know how to build software that they actually need to ship to customer laptops?
October 22, 2024 at 5:53 PM