VictoriaMetrics Observability Blog

Blog posts

• 

  • Open Source Tech
  • VictoriaMetrics

  The Data Race That Wasn't a Bug (and the One That Was)

  You turn on Go’s race detector and get a data race inside net/http that makes no sense at first. This post digs into how Go’s HTTP client really sends a request body, why it can keep reading your buffer after Client.Do has returned, why we decided that race is harmless in vmagent, and why the same behavior could corrupt data in vmauth, so we fixed it there.

• 

  • Go @ VictoriaMetrics
  • Open Source Tech

  How Go’s Built-in Map Works with Swiss Tables

  Go 1.24 replaced the built-in map’s bucket-based runtime with Swiss Tables. This article explains groups, control bytes, H1 and H2, probing, table growth, directories, deletion, load factor, and the experimental split-group layout.

• 

  • Go @ VictoriaMetrics
  • Open Source Tech

  Understanding Go's sync.Map from API to Hash Trie

  Go’s sync.Map now uses a hash trie. This article builds the implementation from a normal map, then explains lock-free reads, fine-grained writer locking, collisions, retries, deletion, Clear, Range, and the trade-offs against map plus RWMutex.

• 

  • Go

  Go 1.27 interactive tour

  An interactive tour of what’s new in Go 1.27: every notable language, runtime, and standard library change, with short runnable examples you can edit and run in the browser.

• 

  • Go @ VictoriaMetrics
  • Open Source Tech

  Monotonic and Wall Clock Time in the Go time package

  Operating systems expose a wall clock that can leap or slew with NTP and a monotonic clock that never runs backward. In Go, only time.Now (might) carries both readings, while values from time.Parse, time.Date, etc., are wall-clock-only—so naïve equality checks or time.Since on those can mislead when the system clock shifts.

• 

  • Go @ VictoriaMetrics
  • Open Source Tech

  Go synctest: Solving Flaky Tests

  Traditional concurrent Go tests can be flaky due to non-deterministic scheduler behavior and timing. Go 1.24’s experimental synctest feature provides deterministic testing by running goroutines in isolated ‘bubbles’ where a synthetic clock only advances when all internally managed goroutines are durably blocked.

• 

  • Go @ VictoriaMetrics
  • Open Source Tech

  Graceful Shutdown in Go: Practical Patterns

  Go applications can implement graceful shutdown by handling termination signals (SIGTERM, SIGINT) via os/signal or signal.NotifyContext. Shutdown must complete within a specified timeout (e.g., Kubernetes’ terminationGracePeriodSeconds)…

• 

  • Open Source Tech
  • Monitoring

  Prometheus Monitoring: Functions, Subqueries, Operators, and Modifiers

  Prometheus querying involves multiple moving parts: functions, operators, modifiers, and subqueries. Small differences in how vectors are handled or rollups are applied can lead to completely different outcomes or errors. MetricsQL extends PromQL by handling these edge cases more gracefully with features like default rollups.

• 

  • Open Source Tech
  • Kubernetes
  • Go @ VictoriaMetrics
  • Monitoring

  Container CPU Requests & Limits Explained with GOMAXPROCS Tuning

  When running Go apps in Kubernetes, default CPU thread scheduling can conflict with cgroup CPU limits. The runtime sees all host CPUs, but the container may only be allowed a fraction of one. This often leads to early throttling. Properly configuring GOMAXPROCS avoids this waste and improves stability.