Observability of your app in production

Just shipped your app to the Play Store for the first time? You'll want to make sure you have visibility into what is going on in the wild. We'll go through an overview of some of the most useful and often free tools, such as Play Store vitals; Firebase reporting; logging, and performance monitoring.
You'll leave this talk with the knowledge to help improve the quality of your app and avoid fire fighting in production.

Transcript

1. September 1st, 2022 Observability of your app in production Kurt Nelson Senior Software Engineer, Pinterest @kurtisnelson
2. 2. When do I set it up? 1. What is it? Observing Today’s Talk 3. How do I do it? 4. Now what?
3. Observability is the ability to see what is happening in production on user devices with your binary Unstructured logging, crash reporting, performance metrics, and analytics are the common forms.
4. As soon as an app is in the public’s hands, it should have basic, manual, observability. Don’t rely on just a QA process.
5. Google provides the basics, for free! Bigger companies often roll at least part of their observability stack, often analytics so that it can tie into other data pipelines.
6. Bread & Butter Traditional Logging Think LogCats, non-fatal stack traces, unexpected but recoverable behavior all sent as strings. Crash Reporting The crash itself is a structured log; a good system will collate these with relevant unstructured logs. Clustering and attaching of metadata pertaining to the build is also performed. Structured Logging This is a catch-all term, and is often speci f ic to the business needs of your application. Think performance data, analytic events, or ad impression data.
7. Think JSON without a well-de f ined schema, or LogCat output on the terminal. Extra work required for machine consumption. Structured Logging Think something represented by a data class, protocol buffer or thrift de f inition. Easily consumed by machines. Unstructured Logging
8. Do you use structured logging?
9. Unstructured data is always better than no data. (Break out Grep in case of emergency)
10. Actions to take based on the data can be determined on the f ly until process is achieved. The process resulting from observability is second. Getting the data f lowing into systems is the important part.
11. Toasting your Bread & Butter Experimentation Observing the known unknowns Alerting Observing the unknown unknowns Feature Flags/Circuit Breakers Tools to trigger when you observe something wrong
12. When do we implement? Observability
13. In a microservice architecture, this comes out of the box. But as usual, your mobile app is probably the only one.
14. Google Play Console Vitals Crashes, ANRs, and glaring performance issues
15. When do I need more?
16. When do I need more? Communicate to business stakeholders what you are able to observe!
17. Align with success criteria Communicate to business stakeholders what you are able to observe!
18. How can I implement observability? Observability
21. Structured Crash Logging
22. Structured Everything Logging
24. Honorable Mentions Bug Snag A commercial crash reporting system mobile.dev Catch performance issues pre-production Jaeger Tracing With proper backend instrumentation, observe the user f low through the backend too.
25. Big Tech likes to do these in house.
26. Now what? Observability
27. Relax. Next time there is an outage, you’ll be equipped. But what if you can avoid the outage in the f irst place?
28. Google came up with SRE to describe the backend engineers treating operations as code. For mobile, the service is the app itself. We can crib some of their best practices. See https://sre.google/sre-book for a deep dive!
29. 1. Alert when things get strange but still seem to work. 2. Guard changes and automatically revert them based on observations. Once we have observability into our app we can…
31. September 1st, 2022 Observability of your app in production Kurt Nelson Senior Software Engineer, Pinterest @kurtisnelson