Thread by @RaeKnowler

.@roidelapluie is talking about Monitoring as an entry point for collaboration. Here’s the CAMS principles of DevOps: we’re going to talk about all of them. #DevOpsDays

Okay, you have your dashboard with green and red lights. And green, red, yellow, red, green? What does that all mean? It doesn’t tell us much about what’s really going on. Maybe the light is green but there’s a traffic jam...

Observability is about the ability to ‘be’ inside the system, with high-quality data, delivered rapidly. #devopsdays

There are lots of different kinds of metrics to record and monitor. Important to remember: metrics do not represent problems. They represent state. You can alert on them (@GianArb I guess this is what you meant). #devopsdays @roidelapluie

Automate your monitoring so that you can be certain it reflects the state of reality. @roidelapluie #DevOpsDays

.@roidelapluie recommends Prometheus: it’s cloud native, scales, has dynamic config, has various services that can be used for different projects (so you can use Prometheus for everything, instead of a different monitoring stack for each). #DevOpsDays

There are also exporters from most languages and systems to Prometheus. However, it doesn’t visualise data... for that, connect up Grafana. (Prometheus recommends this too.) @roidelapluie #devopsdays

We can now see metrics from our systems, cool, but: CPU usage != money. We need business metrics: metrics that represent what really matters to the customers. These require a *cross-team* understanding of the business. #devopsdays @roidelapluie

Once we’ve got these on the central monitoring solution, the different teams can all see them, gaining more useful understanding of what the system is doing. 👍 #DevOpsDays @roidelapluie

You can even use SQL exporters to get data from the database into Prometheus. This requires cross-team communication (good) and you get fine-grained, relevant data. Don’t export so much you might as well build your dashboard on top of the DB! #devopsdays @roidelapluie

.@roidelapluie sets up business dashboards in two parts. On top: some graphs—they try to keep the same selection across projects, for quick recognition—and customer/business metrics. Below and hidden by default: more technical metrics. Don’t focus on those. #DevOpsDays

Keeping your dashboards consistent between projects and colour-coding the info consistently is not just important cross-team: also think of newbies, the on-call team, all the people who take a look during an incident. Help everyone understand quickly! #devopsdays

During a crisis: Alert the right people, including partners of on-call people if necessary. Set up a situation room: both physically (everyone sits together) and a unique channel for discussions, closed after the incident is over. #devopsdays @roidelapluie

@threadreaderapp Unroll, por favor.