Engineering writeups and incident post-mortems for #highload.
Articles: 6In-depth engineering analysis: Multi-agent orchestration: tool call streaming and token latency optimization. Practical diagnostics, logs and production configuration.
In-depth engineering analysis: PostgreSQL connection starvation: PgBouncer transaction mode pitfalls with prepared statements. Practical diagnostics, logs and production configuration.
In-depth engineering analysis: Linux sysctl tuning for 100k concurrent WebSocket connections on a single node. Practical diagnostics, logs and production configuration.
In-depth engineering analysis: Kafka consumer rebalance storm: why rolling deployments caused cascading partition lags. Practical diagnostics, logs and production configuration.
How omitting keepalive pools towards backends results in ephemeral port exhaustion, TIME_WAIT socket buildup, and a 4x latency penalty at high load.
Friday evening incident post-mortem: how a single unbatched 2M row UPDATE paralyzed replication replica via doublewrite buffer contention and sync_binlog=1.
