Performance: Avoiding Pitfalls, Diagnosing Issues & Scaling for Growth
summary
Shows concrete results: a Jenkins master with 170,000 jobs is possible but fragile; changing JVM flags, upgrading JDK, switching to G1 garbage collector, disabling explicit GC calls and setting heap to 4GB+ with min=max cuts pauses and yields a 3,500% performance increase in customer cases. Recommends tools GC Easy and FastThread, jstack/jcmd, support bundles, SSD storage, NFS 4.1, rotating build history, and horizontal scaling once heaps approach ~16 GB.
AI-written summary · about
chapters (35)
- 0:00Introduction
- 4:02Good Morning
- 6:51Basic JVM Architecture
- 7:12Java 8 Architecture
- 8:38JVM Administration
- 9:20Magic Arguments
- 10:15Macro Metrics
- 10:25Micro Metrics
- 10:54System Administration
- 11:36What its like to administer a Jenkins server
- 12:27Our experience with Jenkins performance
- 12:56Java performance tuning skills
- 14:54Initial set up best practices
- 20:01Agent performance
- 23:06Monitoring Jenkins
- 23:43Initial Diagnosis of Issues
- 26:32Slow Requests
- 27:35Build History
- 28:02GC Logs
- 29:01Thread Dump Analysis
- 29:44Garbage Collection
- 31:09Garbage Collection Events
- 32:03Young vs Old Generation
- 33:01CGC
- 34:06Thread Dump
- 35:56Threaddump Analysis
- 37:52Jenkins is Slow
- 40:14Garbage Collection Analysis
- 41:41JVM Arguments
- 43:37Garbage Collection Cycles
- 44:14High Availability
- 46:51Scaling Out
- 48:05Scaling horizontally
- 49:03Resources
- 49:44Booth
related talks
01h 01mOpening Keynote
42mSeven Habits of Highly Effective Jenkins Users- 42mSeven Habits of Highly Effective Jenkins Users
55mContinuous Integration on Steroids
38mOperating Jenkins at Scale
56mOpening Keynote
01h 02mTools for simplified large scale systems
28mGrowing Pains - Scaling Jenkins in the Federal Sector