Advertisement ยท 728 ร— 90

Posts by Ben Wilson

New Opus model, how are people finding it in the last 30 mins?

5 days ago 0 0 0 0

Diffsitter is also a good choice. Integrates must tree sitter flavours

2 weeks ago 0 0 1 0

There were a couple of slop-bots which kept appearing in the bsky.app/profile/devo... feed. I've blocked them and cleaned the feed up a little.

Unfortunate because their content was interesting - but AI generated opinions, which doesn't fit well to the feed.

3 weeks ago 0 0 0 0

Oh, well I got one point on taint bingo!

I actually used the word today as well in a none k8s context. I tainted that flaky VM!

3 months ago 2 0 0 0

No they literally make buttplug software ๐Ÿ˜‚ But I'll be honest, kubernetes was my first guess

3 months ago 0 0 1 0

Interested. If you see anything please ping me with what you find.

3 months ago 1 0 0 0

Referring to this if you're wondering - opentelemetry.io/docs/specs/o...

4 months ago 1 0 0 0

Oh really? I guess it's a bit of stockholm syndrome for me ๐Ÿ˜‚ Honestly I think both do "traditional observability" very well. But you're right, datadog probably does it best.

Interestingly I've noticed the OTel spec is picking up on some things which will enable other vendors to behave like datadog

4 months ago 1 0 1 0

I still regress to dashboards ๐Ÿ˜‚ But that's out of comfort not necessity. I will say, Honeycomb has nailed the needle in the haystack searching, you really don't need dashboards (unless you count a query page as one? ๐Ÿคทโ€โ™‚๏ธ)

4 months ago 1 0 0 0
Advertisement

I see dashboards as casting a wide net, it's the "I vaguely know what the problem might be" versus an SLO which is much more narrow.

Even with SLO's, you'll still have stuff you forget or don't know to cover and revert to needle in haystack searching.

4 months ago 1 0 1 0

Absolutely agree btw, we did an eval with Honeycomb and I got to see what Martin is referring to firsthand. But I also come from the background of LGTM stack.

4 months ago 1 0 2 0

Interesting! Is the source IP vaeem owned or on their docs anywhere?

4 months ago 0 0 1 0

At a guess, you have Veeam running in the cluster and probably have some TLS interception going on and getting spammed - hence maybe you noticing it? As to why it's got such a garbage hostname the feature branch thing is just a guess lol

4 months ago 0 0 1 0

Honestly the name looks like a feature branch. As to why it's hitting your cluster in particular, do you work for Veeam or use Veeam? If either of those, there's your answer. It's broken ๐Ÿ˜‚

4 months ago 0 0 1 0

Sure! There's actually additional services which let you do basic regex based feeds yourself. It's not part of the app and is rudimentary, but it's still something usable.

I think popular feeds get monetised. You can do pinned posts etc. That's the reward for running a popular feed ๐Ÿ˜‚

4 months ago 1 0 0 0

The cost comes from getting the post ID's. So a small VM (I use OVH cloud) is enough to process the stream. It's a simple websocket with IIRC their own ZSTD dictionary for optimal compression. There's no bandwidth overages or anything for the VM thankfully.

Processing can be regex/LLM etc.

4 months ago 0 0 1 0

Minimal, so the feed itself is just a skeleton endpoint, so when a request comes in, all you return is an array of post ID's. I didn't implement pagination but you can do that too ๐Ÿ˜‚

4 months ago 0 0 1 0
Advertisement

Hah it's all good, very cheap VM for ingesting the data and a cloudflare subscription. I can make use of it for other things as well

4 months ago 0 0 1 0

Aww man I just logged into the account, considering that I might wind down the feed. But it's low maintenance and low cost so seeing this I'll keep it going ๐Ÿซถ

4 months ago 1 0 1 0

Yeah 100%. I run on a cheap VPS which then ships data to cloudflare workers and data is served off cloudflare workers too.

$5 VPS on ovh handles it with no overages on bandwidth

5 months ago 2 0 0 0

Yeah most the core product team lost their job at GitHub and instead they're focusing on AI. The statement doesn't exactly hide the AI part but it doesn't explicitly mention it. I'm ready to start evaluating different providers for CI. It's really annoying me how under invested it is

5 months ago 1 0 0 0

Call me a tool please Kelsey, it would be an honour!
I'll pop it on my resume

6 months ago 1 0 0 0

Most "DevOps engineers" don't actually know what DevOps or DevOps practitioning is. Probably not that hot, just inflammatory ๐Ÿ˜…

6 months ago 1 0 0 0

Hah that's neat! I've always wanted to do an interactive talk. I know when I get the time and lack of severely crippling anxiety to talk in front of a room of strangers that's the kind of talk I'll do. Thanks for sharing man!

6 months ago 1 0 0 0

Hah nice interactivity!

6 months ago 0 0 1 0

Some callouts for SI already. I found it okay, it was only AWS at the time I looked but hopefully it's changed?

Terraform is cool still. Without drama attached you might prefer OpenTofu the fork, but license likely doesn't matter and the features are so close it's a bit of a non starter convo.

6 months ago 1 0 0 0

Current: frustrated

7 months ago 1 0 0 0
Advertisement

Hah love that. Just so long as you don't explain a firewall being like a Carburetor ๐Ÿคฃ

That's cool, please update us with how you get on! Love to celebrate the wins

7 months ago 1 0 0 0

Nice! What was the question you handled best and what do you think was the weakest answer?

7 months ago 0 0 0 0

Now I do!

7 months ago 1 0 0 0