Common Causes of 100% CPU Usage in Java Backend Services

CPU usage reaching 100% is a painful, often non‑deterministic problem that can appear after a system has been running for a while.

1. Excessive data fetching – In a restaurant ordering system the consumer mistakenly retrieved the full list of dishes instead of incremental updates. The massive payload triggered frequent full gc cycles, causing the CPU to spike.

2. Kafka auto‑acknowledge – Using Kafka’s automatic commit reduced code but, as message volume grew, the consumer re‑processed messages and drove CPU to 100%. Switching to manual acknowledgment eliminated the issue.

3. Infinite loops – Classic while / for loops or uncontrolled recursion keep a register or recursion depth counter alive, continuously consuming CPU. In JDK 1.7 a HashMap put under multithreading could also enter a dead loop.

4. Multithreaded data import – Replacing a single‑threaded Excel import with a thread‑pool improved speed, but a large number of threads caused excessive context switching, raising CPU usage.

5. Bulk file synchronization – Regenerating static HTML sites for many games and copying all files to a web server at once overloaded the application server’s CPU.

6. Deadlocks – Improper lock ordering (e.g., using synchronized or Lock) can cause two threads to wait on each other. The following thread dump illustrates a waiting thread:

"pool-4-thread-1" prio=10 tid=0x00007f27bc11a000 nid=0x2ae9 waiting on condition [...] java.lang.Thread.State: WAITING (parking)

7. Inefficient regular expressions – Java’s NFA engine performs backtracking. A poorly written pattern such as

^([hH][tT]{2}[pP]:(//|[hH][tT]{2}[pP][sS]://))((\w|\-)+\.)+([A-Za-z0-9-~/])+$

can cause prolonged CPU consumption.

8. Heavy real‑time calculations – Computing discounts or aggregating large data sets under high concurrency can keep the CPU busy for extended periods.