Notes on seeing the future
Engineering, product and guides from the Foreseer team.
Predictive vs reactive monitoring: why thresholds keep failing you
Static thresholds alert you to outages you are already having. The engineering case for forecasting — with the math, the failure modes, and where prediction honestly does not help.
Anatomy of an alert worth waking up for
Symptom, cause, blast radius, fix — the four parts that separate an alert an engineer acts on from one they mute, and why most tooling ships only the first.
How to monitor a Redis cluster (and forecast OOM before it drops writes)
The Redis metrics that actually predict failure — memory vs maxmemory, eviction policy, fragmentation, replication lag — and how to watch a whole cluster from one agent.
Predicting Elasticsearch heap pressure before nodes drop out
Old-gen occupancy, GC time, circuit breakers and shard count — the leading signals of a heap crisis, and how to forecast it instead of reacting to a node leaving the cluster.
More posts coming soon.