Measuring Correctness with Synthetic Checks

#96 - Practical Guide to Implementing SRE and SLOs - Alex Hidalgo

Topic 6 of 11

Published Jul 11, 2022 · Clip 37:33–42:02 (4:29)

Alex explains that data correctness measurement depends on the service. He describes using synthetic checks: at Google he built a crafted log with a special tag that flowed through the entire pipeline (producers, Kafka, Logstash, Elasticsearch, Kibana), verifying availability, latency, errors, correctness, and freshness — covering five services with one measurement running only a few times per minute. He notes you don't need as much resolution as you think, and that SLIs can also be built from real user traffic via advanced tracing, which is possible but a lot of work — be thoughtful and do what works, even if that's just the golden signals.