Works with what you already run

Cluster health, JVM heap pressure, GC pauses, search and index throughput, thread-pool rejections and shard allocation.

✦ We forecast heap saturation and flag unassigned-shard risk before the cluster goes yellow — not after.

Connection saturation, cache-hit ratio, lock contention, long transactions, table bloat, autovacuum lag and txid wraparound.

✦ Per-node replication topology with lag — plus a txid-wraparound forecast that warns days before an emergency vacuum.

Connections, InnoDB buffer-pool hit ratio, slow queries, deadlocks, temp-disk tables and replica thread health.

✦ We map the full replication topology and forecast capacity, so a lagging or stopped replica surfaces immediately.

Memory and fragmentation, eviction and hit rate, connected clients, persistence lag, and replica health.

✦ We project time-to-OOM against your maxmemory policy, so a noeviction instance never silently starts rejecting writes.

Per-pipeline event throughput, persistent queue depth, worker back-pressure and JVM garbage collection.

✦ We watch queue depth slope, so a stalled output is caught while it is still draining — not when the disk fills.

Per-container CPU throttling, memory pressure and OOM kills, restart loops, blkio and network errors.

✦ OOM-kill and restart-loop detection per container catches a crash-looping sidecar before it takes the host down.

Active connections, request and error rates, connection saturation and upstream state imbalance.

✦ Peak-capacity forecasting projects when connection limits will be hit under your traffic trend.

Worker and scoreboard saturation, request spikes, and busy-vs-idle balance from mod_status.

✦ A worker-exhaustion forecast warns before MaxRequestWorkers starts queuing requests.

Heap and metaspace pressure, GC pause time and frequency, thread states, deadlock detection and class-loader leaks.

✦ Heap-regression forecasting estimates the OOM ETA, and live Jolokia deadlock detection catches lock cycles.

Pool saturation, max-children hits, listen-queue backlog and slow-request build-up.

✦ Worker-saturation forecasting flags when a pool is about to start queuing requests.

Watch any named process: liveness, restarts and restart loops, and RSS / CPU / file-descriptor pressure against its limits.

✦ An RSS memory-exhaustion forecast works for anything you run — even services without a dedicated analyzer yet.

Filesystem fill, inode exhaustion, CPU saturation, memory pressure, load average and IOPS — on every host.

✦ Linear and seasonal fill forecasting tells you the date a volume runs out, with days of lead time.