How would you diagnose HikariCP connection pool exhaustion?

Uses Hikari metrics, thread evidence, database activity, transaction scope, and query behavior to separate leaks from slow or excessive work.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I first confirm the symptom: active connections near the pool maximum, pending acquisition growing, and acquisition timeouts. Then I correlate the start with traffic, deployment, query latency, and database health. Pool exhaustion can come from leaked connections, long transactions, slow or blocked queries, too much parallel database work, or a pool that is too small for a changed workload.

Full answer below

Practice focus

Pool pending and active metrics, acquisition timeout, leak suspicion, thread dumps, database sessions, transactions, and root cause.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
Pool exhaustion means demand for connections exceeds their return rate; a leak is only one possible cause.
I first confirm the symptom: active connections near the pool maximum, pending acquisition growing, and acquisition timeouts. Then I correlate the start with traffic, deployment, query latency, and database health. Pool exhaustion can come from leaked connections, long transactions, slow or blocked queries, too much parallel database work, or a pool that is too small for a changed workload.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
I inspect thread dumps for requests waiting on Hikari and for threads holding transactions while doing non-database work. Database activity and lock views show long-running statements, blockers, and sessions by application instance. With Spring Data and transaction management, ordinary managed access should return connections, so I focus on manual JDBC handling, streams or cursors left open, remote calls inside @Transactional, and new fan-out loops. Leak detection can provide clues but is not a substitute for correlation.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Assuming every exhaustion incident is a leak.
Restarting before collecting thread and database evidence.
Increasing maximumPoolSize as the first fix.

How would you diagnose HikariCP connection pool exhaustion?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section

Tech stacks

Skill areas

Suggest an improvement

Found this answer useful?