A production broker runs out of disk space. What happens and how do you recover?

What happens: 1) Broker stops accepting writes (log flush fails). 2) Log replication to this broker stops. 3) If this broker has partition leaders, those partitions become unavailable for writes. 4) If ISR drops below minISR due to this, acks=all writes are rejected cluster-wide for affected topics. Recovery: 1) Immediate: expand disk (EBS resize, add volume, mount new path). 2) Reduce retention: decrease retention.ms or retention.bytes for large topics. 3) Temporarily reduce replication traffic: throttle reassignment. 4) Delete unused topics. 5) Move partition leadership away from full broker using preferred leader election. 6) Add a new broker to rebalance partitions. Prevention: alert at 70% disk usage.

Expected key concepts: disk space, log flush fails, partition unavailable, ISR, minISR, expand disk, reduce retention, throttle, delete topics, preferred leader election, add broker, alert 70%