MartinDotNet

It's cheap if you win at the casino though ;)

AWS is a billing system masquerading as a cloud provider.

AI coding assistants with observability? Honeycomb’s integrating Cursor with MCP so AI can query traces, metrics & SLOs—right from your IDE. Faster, more accurate debugging is here.

Read more: buff.ly/So2KoEc

#honeycomb #observability #AI

* **Full OpenTelemetry Integration:** Deep observability into your message flows with comprehensive **OpenTelemetry** support. Trace messages across transports, outboxes, inboxes, and even claim-check patterns to diagnose performance issues and understand the full lifecycle of messages.

And the best request queuing system in the world.

I have about 1507 systems in my house, and I crash them daily.

If you're interested in Event Driven Architecture, Observability and Beards, this is definitely the talk for you!

We tried to demystify what options you have to make instrumentation work for you when Asynchronous communication is the norm in your application.

Let us know whether it helped!

YouTube video by NDC Conferences Modern Observability and Event Driven Architectures - Martin Thwaites & Ian Cooper - NDC Oslo 2025

This was a tough talk to get right, and I'm not convinced @icooper.bsky.social and I got everything. At the very least, I hope it made you think about how you instrument your Event Driven Apps.

www.youtube.com/watch?v=J7K5...

At London O11y Day, @Liz Fong-Jones said it best:

“Honeycomb is about raising the ceiling and the floor.”

Catch her keynote in SF on Sept 11—how why AI can empower SRE teams and make observability more accessible for all.

Register: buff.ly/j72AYIH

I'd argue that this is the wrong comparison, if people are sleep walking into just adding everything, that's a bigger problem.

If you think it's expensive to add good observability, wait until you see the cost of bad observability. It might not show up as a line item, but it's impact is there.

@jessitron.bsky.social asked someone to help facilitate a DevRel planning meeting a few weeks back. He came out of the meeting saying "how on earth do you lot get anything done".

Putting all the devrels in a room was not a good decision when the plan is to achieve something concrete.

Manager: "Corporate would like to increase the open rate of our emails this quarter"
Product Manager: "I got you bruv"

It's amazing that after a breach, there's a tonne of money for info sec too, but after some downtime there's still no money to level up observability.

I'm not talking just about tooling though, it's about people, infrastructure resources, and general appetite to spend money.

This is ultimately because the exporter extensions also add a Head sampler, which is very naughty, and not something we'd allow in the main contrib repo.

If your users had choice, would they voluntarily choose you and your team? If not, it might be time to take a long hard look at how you're tackling observability.

This isn't new, internal enablement teams have been thinking like this for decades, but it seems like it's new to Observability.

This is really about producing tools and processes that help people, then you don't need to force users to use them. If you're forcing people to use the tool, it's not the right tool.

Listen to your users, give them what they want, not what's easier for you.

This is a great example of thinking like a product team. If your users (engineers in your organization) wouldn't recommend your team/tool, you're not doing your job.

I'd guess that you're not providing the engineers what they need, or helping them use those tools.

This is Observability culture, and strategy, and looking at how the people and tools they use allow them to help their users.

One company I'm working with right now is looking at NPS (Net Promoter Score) for their observability team

This measures how many of your users would recommend your service

The teams that I love working with are those that ask questions like "how can we help the engineers make this better", not "how can we get them their data".

These teams are thinking in terms of how they deliver the best product that helps their users, not give them more dashboards.

Those "Observability teams" are just a cost right?

What we've seen over the last few years is that those teams that do REALLY well are the ones that actually treat themselves as a product team. The leaders have a product mindset.

They're thinking about how they add value, not mitigate it's loss.

So many times I've heard this stuff, and it just under-values your platform and SRE team. Diminishing their role to restarting servers, and scaling things up.

Reducing server sizing, reducing scaling, can save infrastructure costs, decreasing response times, and errors will increase sales.

An info sec incident might have just cost $5m, but your ongoing issues with production causing slowness, errors, and general bad user experience are costing you way more than that through attrition.

it could be "we need more people so we can fix bugs" or "we need better tools to investigate"

It's amazing that after a breach, there's a tonne of money for info sec too, but after some downtime there's still no money to level up observability.

I'm not talking just about tooling though, it's about people, infrastructure resources, and general appetite to spend money.

Yorkshire is indisputably God's Own County, because it's the only one mentioned in the Bible.

"Three men came from the East Riding on camels"

How was it moving the playlists?

Does anyone know a lot about cloudflare's domain modelling? I'm specifically looking at potentially pro pricing, but I have 20 domains that are all pointing at the same place via CNAME's.

I'm trying to workout how I can not pay 20x $20/month to get pro for them all. I basically want analytics.

Same, my workshop is focused on supporting production apps, and we use the Aspire Dashboard to help get the telemetry right locally.

That post is very "whatever you do, don't press the button"... there is only one outcome.

I can totally imagine what a film for the middle 80s will portray, but... now I kinda want to see it.

Aspire as a control plane I can understand, with the dashboard, docker etc.

ServiceDefaults however, thats just a project.

We need to stop making the Aspire ecosystem seem like an all or nothing.

Aspire doesn't give you telemetry for "free" it sets up an otel config, which is ~10 lines of code

I live!! Just getting back on the social horse. This will be my new home for posting about all things #opentelemetry, plus some of the filmmaking I've been up to. Stay tuned!