Cloud Observability Setup Choice

[xd-Park3rr]

https://grafana.com/integrations/cloud-monitoring-microsoft-azure/

Hey @tjementum,

Quick question, did you ever consider using Grafana Cloud for Azure observability instead of Azure Application Insights / Log Analytics in PlatformPlatform?

Not asking because I think you should change your repo at all. I’m asking because I’m considering doing that in my own implementation to keep long-term analytics/logging costs lower, and I was curious if you’d already looked into it or had a reason for sticking with the native Azure setup.

My father suggested it to me, since he works with a very similar Azure stack. He said Azure analytics never really worked out well for them, and they eventually switched to Grafana because it gave them a massive reduction in costs.

[tjementum]

Yes... I considered. BUT decided not to.

Honestly, I have never really used Grafana. It looks nice, but I don't know the pros and cons. What I do know is that Application Insights works, and I have had enormous success with it. Like all data though, it requires discipline, which I claim is built into PlatformPlatform.

Application Insights does not have the best UX, but for collecting data it's second to none, because everything is integrated, not stitched together. Azure Log Analytics gives you one place where you have your usage data (requests, page views, exceptions, dependency calls, metrics, custom events, browser timings) and your operation data (CPU, memory and other hardware metrics, uptime). Try to set up the Application Insights MCP in PlatformPlatform next time you have a problem and ask it to investigate a bug.

With the new BackOffice I'm one prompt away from getting a perfect dashboard where I can not only see the above mentioned data, but I will also be able to look at individual users or tenants and see what exceptions they have. I will be able to query the system and ask things like: what is the average response time for customers on the premium plan that have the xxx feature flag enabled?

On cost, yes Log Analytics can get expensive at scale, but you can tune sampling and retention to manage that, and for me the integration value outweighs the savings of running a separate stack.

And then there is the compliance angle, which I think is the strongest argument: if you start stitching together your monitoring with Sentry, Grafana, Prometheus, Datadog, Metabase, or whatever else, you need to synchronize them. Plus you also need to update your Data Processing Agreement (DPA) and inform all customers every time you add a new tool.