Date and Time: 2026-10-20 12:30p ET

What is a good way to measure the reliability of dozens of microservices?
* https://www.alex-hidalgo.com/the-slo-book
* SLOs (service level objectives) on the user-facing capabilities.
* Set an error budget and fix it only when the error budget is exceeded. How much are you willing to invest in reliability?
* SLOs assume you have observability that is clean (enough).
* Avoid the trap of creating SLOs around metrics just because they are easily available (eg: machine level metrics like CPU, disk, network IO)
* Do watch metrics for changes and correlate to trailing user experiences, but don't alert on them.

Observability ontology (Netflix recent blog post) https://netflixtechblog.com/ontology-driven-observability-building-an-end-to-end-knowledge-graph-at-netflix-scale-421392cf5adf
* rare outside of hyperscalers (or maybe just Netflix)
* different groups care about different slices and sets of aspects in a combinatorial way
* useful for setting context for meetings between humans

What are the best tools and practices for observability in production systems? How do you decide which tools to use for logs, metrics, tracing, and alerting?
* No single tool is "best." You probably want to ensure your agents output Open Telemetry to keep you vendor agnostic
* Start with what is already available perhaps from your cloud provider to see what matches your philosophy
* Datadog capable but expensive, opinionated
* Grafana lets you make the dashboards, but you need to maintain
* Engineering tradeoffs just like writing code. Write down your business-specific requirements and constraints and values in order to make a comparison for you.
* Ensure the vendor's ontology matches your business's structure or at least is compatible to support your needs
* If you ask an LLM, then hopefully you've defined your ontology for it, otherwise you might accidentally compound the problem into a telephone game