Hello Community
We're experiencing a significant memory leak issue with our production
Tomcat setup and would appreciate guidance on best practices and potential
fixes.
Environment Details
-
Tomcat Version: Apache Tomcat 7.0.41
-
Heap Size: 6GB (max)
-
Number of WAR files: 7
-
Application Type: Java & Spring Boot
-
Deployment Mode: hot-updates using symlinked WAR files
-
OS: Linux
-
JVM: OpenJDK 8 (1.8.0_412)
Current Deployment Approach
We're using a symlink-based deployment strategy:
1.
Actual Tomcat webapps path: */path/to/tomcat/webapps/*
2.
Git repository containing WAR files: */path/to/repo/warFiles/*
3.
Symlinks created from Git repo to Tomcat webapps directory
4.
Deployment process: *git pull* to update WAR files, no Tomcat restart
5.
*server.xml* configuration:
<Host name="localhost" appBase="webapps"
unpackWARs="true" autoDeploy="true">
Problem Description
Over the past 5 months (since the last Tomcat restart), we've observed a
progressive memory leak with the following pattern:
Class Count Growth (Staircase Pattern):
-
Initial: 50k-56k classes
-
After each git pull/deployment: Incremental increase
-
Peak: 250k classes
-
After restart: Back to 50k-55k classes
Performance Degradation:
Metric
Before Restart
After Restart
Class Count
250k
55k
Max GC Count
56
46
Avg GC Count
10
6
Max GC Time
2000ms
963ms
Avg GC Time
400ms
215ms
Avg JVM Threads
1.44k
277
Monitoring
We've been monitoring this issue through Zabbix, and the graphs clearly
show the progressive degradation over time (includes incident &
after-incident pattern as well):
-
Class count growth over the months (showing staircase pattern from 50k
to 250k)
[image: Screenshot from 2025-11-24 18-58-46.png]
-
GC activity trends (count and time)
[image: Screenshot from 2025-11-24 19-08-21.png]
-
Memory/Heap usage over time
[image: Screenshot from 2025-11-24 19-13-01.png]
-
Thread count growth
[image: Screenshot from 2025-11-24 19-13-52.png]
Incident: After a recent git pull, Tomcat experienced an OutOfMemoryError,
generated heap dump files, and crashed. The system was unresponsive until
we performed a restart.
Questions for the Community
1.
Is our symlink-based deployment approach with autoDeploy="true" causing
classloader leaks??
2.
What is the recommended deployment strategy for production environments
to avoid memory leaks when updating WAR files??
3.
Are there known issues with Tomcat 7.0.41 related to classloader leaks
during redeployment??
4.
What monitoring or preventive measures can we implement to detect and
mitigate this issue before it causes production outages??
We suspect the combination of symlinked WAR files, autoDeploy="true", and
no-restart deployments is preventing proper cleanup of old classloaders,
but we'd like to understand the root cause and implement a robust solution.
Any insights, best practices, or references to similar issues would be
greatly appreciated.
Thank you for your time and assistance.
Thanks & Regards
Harshit Sharma
