Alerts: AnkusaBreakerOpen
At least one destination has had its circuit breaker open for 5 minutes. Its deliveries are parked: written back due when the breaker's period ends, attempts not spent. Every other sink is unaffected. The node logs a warning when a breaker opens.
ankusa_dispatch_breakers_open: how many are open on the node now.ankusa_dispatch_breaker_transitions_total{sink,state}: which sink opened (state="open") and how often it has closed again.ankusa_dispatch_attempt_duration_seconds by sink and result: what the failing attempts look like (error, or timeout at dispatch.attempt_timeout_ms).ankusa_queue_oldest_due_age_seconds: the cost of the park, which goes up behind a breaker.A breaker opens after dispatch.breaker_failures (5) consecutive failures, stays open for dispatch.breaker_open_ms (30 s), then doubles per consecutive open up to dispatch.breaker_max_open_ms (5 min). Then one attempt probes.
GET /v1/dlq shows the exact reason string of rows that gave up; the log has the first failures. Typical causes: the destination is down, a credential expired or was rotated, a network policy changed, a certificate expired, the destination is rate limiting us.PUT /v1/tenants/{tenant}/sources/{name}). The next probe closes the breaker; a breaker that last opened long ago re-probes within breaker_max_open_ms (5 min).dispatch.retry.max_elapsed_ms, 24 h) and replay them later.ankusa_dispatch_breakers_open returns to 0, and ankusa_dispatch_breaker_transitions_total{state="closed"} rose for the sink.result="ok" and no error or timeout.ankusa_queue_oldest_due_age_seconds falls: the parked rows are being delivered.Can you improve this documentation?Edit on GitHub
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs |
| ← | Move to previous article |
| → | Move to next article |
| Ctrl+/ | Jump to the search field |