abandoned-routine events dropped [SQL0052]
Edit this pagecode SQL0052 · log event at warn — degraded but self-healing, or a durable data consequence; an operator should learn of it eventually
the log line
As it arrives in your log:
level=WARN msg="abandoned-routine events dropped" code=SQL0052
A consumer instance discarded abandoned/cleared routine events instead of
producing them to __system.metrics; dropped_count carries how many.
Two causes share the line:
- The batch could not land — the flush tick’s produce failed, and the
batch is dropped, not requeued. The
errorattribute carries the cause, usually a database problem other lines will also be reporting. - The in-memory queue hit its cap between flush ticks (no
errorattribute). The cap exists so a storm of abandoned routines cannot pile up unbounded memory or database load; sustained cap drops mean the consumerFunc is hanging past its timeout at a high rate — investigate the consumer (Consumer Timeouts), not the metrics.
The loss is permanent for those events: abandoned-routine snapshots
(Total, Outstanding, self-clear latency) undercount for the window,
and the abandoned-routine alert can miss a fire or a resolve. The
session’s abandoned_count on the stopped line is unaffected — it is
counted in memory, not read back from the stream.
Nothing pasted yet — the queries above show their blanks.
Looking for a different code? Search every thread.