Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

2023 DevOps Is Terrible

abidmoon.hashnode.dev

Post reply on HN
Print view View original

121–126 of 126 posts

Re: 2023 DevOps Is Terrible

#121

Post by fragmede »

tester756 wrote: post #71
KaiserPro wrote: post #51

I think the biggest issue with devops is that it originally meant "socialising" your sysadmins, by getting them to sit with your devs, so that shit didn't get lost because nobody thought to talk to the right team. But then it morphed into "oh lets innovate with infrastructure" but the innovation turned into "lol lets just restart from scratch and ignore history" Anybody who used early k8s can attest to how un product…

Admins wanted to increase their $$ and started calling themselves DevOps And developed some automation to justify that

I know a few old school sysadmins who got left behind because they couldn't upskill to DevOps and consequently lost out on $$, others who were fired after moving everything off prem to AWS because the company no longer needed them. You understand that your theory is a "conspiracy theory", so why even say it? how does that add to the discussion?

Re: 2023 DevOps Is Terrible

#123

Post by tryauuum »

fragmede wrote: post #120
tryauuum wrote: post #96

Earlier quoted context omitted.

god do I hate the "pet" analogy yes this server is special to me. No, I cannot move workloads somewhere else and shoot the server in the head. Because customers are running GPU workloads there, I cannot transparently migrate them

How does the analogy falling flat for you? GPUs are interchangable resources; an H100 is just like any other H100, even if you can't take a snapshot of an intermediate state from one and load it onto another and resume.

I thought about it and I don't hate the analogy. Although usually I'm very against comparing IT stuff with physical stuff because it tends to produce pointless conversations.

What I hate is the attitude of "treating your server is a pet is wrong, you should be able to kill it and replace any time". I feel like people who say it never run anything more complex than an HTTP API.

Re: 2023 DevOps Is Terrible

#124

Post by fragmede »

tryauuum wrote: post #123
fragmede wrote: post #120

Earlier quoted context omitted.

How does the analogy falling flat for you? GPUs are interchangable resources; an H100 is just like any other H100, even if you can't take a snapshot of an intermediate state from one and load it onto another and resume.

I thought about it and I don't hate the analogy. Although usually I'm very against comparing IT stuff with physical stuff because it tends to produce pointless conversations. What I hate is the attitude of "treating your server is a pet is wrong, you should be able to kill it and replace any time". I feel like people who say it never run anything more complex than an HTTP API.

Ah yeah I totally agree. Having pets is fine until it isn't. A single-digit handful is fine. 1,000 bespoke pets, way above Dunbar's number, isn't. Where exactly you run out of mental headroom and should convert is up to you; the point is that there is one and converting over should be planned for in the North Star roadmap of future development.

My (Raspberry Pi) home server is very much a pet, but I'm never going to productionize that thing, so I'm totally fine with it being a pet, despite the practice at work of having no pets, all cattle.

Re: 2023 DevOps Is Terrible

#125

Post by pojzon »

davewritescode wrote: post #110
pojzon wrote: post #39

DevOps is terrible because we want to have a team of fullstack expert engineers that have certs in every tech, but pay them for a single role. First there are not enough ppl with that huge landscape of knowledge. Second you would have to pay a salary of 5 ppl i.e. 500k/y to make ppl willing to spend that much time on work.

There are more people than you think that can slap together some YAML and a Helm chart and deploy something reasonably resilient inside a Kubernetes cluster. I've been around for the cloud migrations in every phase, what we have today with Kubernetes is miles better than the maze of Puppet and other crap we had 20 years ago.

But thats just a tip of an iceberg.

What happens when your Cillium Service Mesh starts misbehaving due to missed network packets ?

Or when you have to setup a resillient backup of a cluster that runs 30 000 pods on 300 nodes ? And upgrade.

Or when you have to provide a telemetry for highly distributed system using 5 languages on landscape spread over frontend and backend with three different db engines and two clouds ?

Kubernetes is easier to start but 10 times as hard to do at scale you didnt even have back then in OnPremise era.

Yes you are right there are a lot of ppl who can write and deploy „Hello World” apps, but thats about it.

Re: 2023 DevOps Is Terrible

#126

Post by rswail »

hgopolis wrote: post #117
rswail wrote: post #103

Earlier quoted context omitted.

Calling a route table "access control" is not useful. It's not access control, it's literally in the name, it is about navigation of the network via routes. As for your comment about the "pocket protector committee" that's just ad hominem attacks on the people that keep the internet running. Saying that NDP is more complicated than ARP/RARP etc because it has a different name is, again, attacking actual needs because…

I design motherboards for network devices; I keep the internet going. Routes are inaccessible without access to them. The underlying machine states to check for access to a route is algorithmically similar to the other checking for user access. It’s unfortunate so many are allowed to work in IT with cliff notes level awareness of how technology works. The layers of indirection you work with are the knobs and buttons…

Well aren't you just the genius holding the internet together.

Of course the underlying algos are similar, because they are table matching and they are also doing either min or max matching depending on the requirements.

They're also sorted for most efficient access. Does that make sorting somehow part of access control?

btw, these days, most of the "internet" is expressed either in programmable hardware or in software. Some of those algorithms may be expressed in raw hardware.

I work with a guy that makes gzip operate in hardware at 10G speeds, I'm aware of the requirements of networking.

I'm also aware of the OSI layers and how they sort of but not quite map to L1-4 of modern networks, which only have those names because of OSI.

So don't insult people here because you do some low level network design.

Post reply on HN