Monitoring

Aeron exposes a rich set of lock-free counters in shared memory. Knowing which ones matter — and what a bad value looks like — is the difference between catching a problem at 2pm and getting paged at 2am.

In this section

• Counters — the shared-memory counter file and the signals worth alerting on.
• aeron-stat — live view of driver and stream counters.
• aeron-stat-diag — automated health checks and what they look for.
• cluster-diag — cluster-specific diagnostics and common pitfalls it surfaces.
• Aeron Insight (Premium) — production observability and diagnostics.

The counters that page you

Signal	What it means	Look for
Back-pressure	Publisher is being throttled by an unready subscriber	Sustained non-zero → consumer can’t keep up
NAKs / retransmits	Packet loss on the wire	Rising rate → network or buffer sizing problem
Loss gaps	Receiver waiting on missing data	Non-trivial gaps → loss + slow recovery
Conductor duty cycle	Driver event loop latency	Spikes → CPU contention / lost core isolation

Monitor before you tune

Counters are also your tuning feedback loop. Change one knob, watch the relevant counter, keep the change only if the signal improves. See Performance Tuning.

Note

For the meaning of each counter at the protocol level, The Aeron Files has a dedicated counters section. Here we focus on which to watch in production and why.