How to Diagnose and Fix High CPU Usage in a Java Web Portal

1. System Check

A portal page timed out; OS load average was over 4 and process 7163 consumed more than 800% CPU.

2. Fault Localization

Identify the high‑CPU thread, examine its stack, and find the problematic database call. top -Hbp 7163 | awk '/java/ && $9>50' Thread ID 16298 converts to hex 3faa.

printf "%x
" 16298
3faa

Use jstack to view the stack for that thread: jstack 7163 | grep "3faa" -A 30 The stack trace reveals a database query.

List MySQL processes and save to a log:

mysql -uroot -p -e "show full processlist" >mysql_full_process.log

Find the most frequently queried table: grep Query mysql_full_process.log The table contains nearly three million rows, causing long query times.

Check the table size and existing indexes:

use databases_name;
select count(1) from table_name;

show create table table_name\G

3. Confirmation and Fix

Data older than one month is not needed; delete it:

delete from table_name where xxxx_time < '2019-07-01 00:00:00' or xxxx_time is null;

Create an index on the device_uuid column: alter table table_name add index (device_uuid); Verify the index was created:

show create table table_name;

4. Result

CPU usage dropped to about 40%; the root cause was a database query without an appropriate index.

5. Other Tips

Adjusting MySQL's innodb_buffer_pool_size (e.g., to 4G) can improve query performance, but adding proper indexes remains the primary optimization method.

Source: https://blog.51cto.com/rongshu/2426712