Skip to main content

Metrics

When metrics_enabled = true (the default), each FileSystemContext spawns a background HeartbeatTask that periodically reports incremental counter deltas to the GooseFS Master via MetricsHeartbeat. Optionally, metrics can also be pushed to a Prometheus Pushgateway.

use std::sync::Arc;
use std::time::Duration;
use goosefs_sdk::config::GoosefsConfig;
use goosefs_sdk::context::FileSystemContext;
use goosefs_sdk::metrics;

#[tokio::main]
async fn main() -> goosefs_sdk::error::Result<()> {
    let mut config = GoosefsConfig::new("127.0.0.1:9200");
    config.metrics_enabled = true;
    config.metrics_heartbeat_interval = Duration::from_secs(10);
    // Optional Pushgateway (requires feature `metrics-pushgateway`)
    // config.metrics_pushgateway_endpoint = Some("http://127.0.0.1:9091".into());

    let ctx: Arc<FileSystemContext> = FileSystemContext::connect(config).await?;

    // Application counters / gauges via the global registry
    let counter = metrics::counter("Client.AppOps");
    counter.inc();

    ctx.close().await?;
    Ok(())
}

Reporting Channels​

  1. Heartbeat — cluster-aggregated metrics to Master (isClusterAggregated=true)
  2. Pushgateway — all metrics via HTTP POST for Prometheus/Grafana

Java-compatible cluster-aggregated counters include Client.BytesReadLocal, Client.BytesWrittenLocal, and Client.BytesWrittenUfs. The Rust client also exports RPC op counts, error counters, latency cumulatives, Client.Cache* page-cache metrics, and Client.MetadataCache* metadata-cache metrics (MetadataCacheHits / Misses / Expirations / Invalidations / NegativeHits / Size / Enabled — see Metadata Cache).

note

Client.GetStatusOps counts Master RPCs only; metadata-cache hits are not counted.

Full catalogue: docs/METRICS.md.

Examples:

cargo run --example metrics_heartbeat
cargo run --example metrics_pushgateway --features metrics-pushgateway