Harpratap Singh

Engineer. Opinions are my own

[Harpratap Singh @theHarpratap]

@oh_smr_ @_etdev Japanese alone isn't enough, you need to combine it with another skill which will open a lot of opportunities. Software + Sales + Japanese customer handling skill is the highest paying category right now for example

[Harpratap Singh @theHarpratap]

@_etdev @oh_smr_ Although I wish they provided more clarity on the after hours oncall work or on weekends and national holidays. Some expect minashi zangyo to cover it while some provide extra comp or days off in lieu.

[Harpratap Singh @theHarpratap]

@ALEXEIMARTOV Yes I did, but I don't hate them. Just a constructive criticism because I still hold a lot of stocks from my days there. I pretty much use all of their big services - Pay, Securities, Ichiba, Card, Books. Mobile is easily the worst product in their portfolio from a consumer PoV

[Harpratap Singh @theHarpratap]

@ALEXEIMARTOV 1 day delivery isn't very useful if I can walk 10 minutes from my home and get literally most of my stuff. They are competing against a well designed walkable city so the western strategies of 1 day deliveries aren't as useful in Japan

[Harpratap Singh @theHarpratap]

@ALEXEIMARTOV Significantly inflating their GMS. Amazon launched same service few months ago so finally we can see if they are actually competitive. But all e-commerce companies in Japan have the same problem - people don't use them enough over here

[Harpratap Singh @theHarpratap]

@KaiHanssen1 I haven't tested it again but not sure if the performance has improved since my benchmarks were done

[Harpratap Singh @theHarpratap]

@KaiHanssen1 The biggest drawback I faced when rolling it out was the learning curve was too steep for someone who just wanted to customize the default dashboards provided in my lib.

[Harpratap Singh @theHarpratap]

@KaiHanssen1 People prefer creating dashboards via UI so they can see if the query is resulting what they hoped it would. So once they have this dashboard json they want a simple translation from JSON > DSL which was very difficult to do with CUE. But with LLMs it shouldn't be a problem

[Harpratap Singh @theHarpratap]

@ahmetb @fiunchinho Not sure if I understand your problem correctly, but config-sync from GKE might be what you're trying to achieve - https://t.co/WMF56Vkgok

[Harpratap Singh @theHarpratap]

@fulmicoton @FrancoisMassot Seems like only premium users can DM you on twitter. But I would be interested

[Harpratap Singh @theHarpratap]

@OFettal @boredabdel @kelseyhightower It isn't necessarily always cheaper though (looking at you GCP) - https://t.co/wKV1uxLOBr

[Harpratap Singh @theHarpratap]

@pranay01 Last time I calculated, the engineering hours required to rip out Datadog instrumentation was so high, you needed to run your otel stack free for 3 years just to break even. And I'm not even including the wasted opportunity cost that could have been used for improving business

[Chris Albon @chrisalbon]

It is embarrassing how much this stupid fucking meme has affected my own thinking and my own mental health.

So, SO many times I’ve been the left person, and someone is explaining to me why I can’t do or think something without reading every book they’ve read, received the degrees they’ve earned, or had the life experiences they’ve had.

It used to bother me.

But every single time now when I find myself in that situation I pause and think, “is this person correct, or are they just the middle person and I should find the right person”

[Karl 🪫 @karlstoney]

Been talking to folks about @IstioMesh Ambient mode recently and a theme I hear a reasonable amount is that scaling Sidecars is difficult. It's hard to explain in a few tweets why this isn't really an issue for us, so I've blogged a bit about it here: https://t.co/PEx5oCxoAu

[Harpratap Singh @theHarpratap]

@aantn @thockin @shakedko @argoproj @unity 2 cpu requests in K8s isn't equal to 2 cores. It's "shares of CPU time". On an 8 core machine with 2 pods of 2 CPU requests you get 4 cores worth on each pod, not 2 cores. Watch this insightful kubecon talk explaining it well - https://t.co/N2o9y9iQLp

[Harpratap Singh @theHarpratap]

@aantn @thockin @shakedko @argoproj @unity It depends on your workload. A lot of web services are very spikey in nature. You can't assume a 2 core Pod will always be using 2 cores. Linux will grant neighbouring pods this spare CPU and you will end up with unpredictable performance. So it's better to have limits

[Harpratap Singh @theHarpratap]

@ahmetb @timothysc Why does your CD have God mode option?

We have mono-repo and have Rego policy set on this mono-repo to allow cluster level resources to be applied only by approval of Platform.