Comprehensive Guide to Java Online Fault Diagnosis: CPU, Memory, Disk, GC, and Network Troubleshooting

Online fault diagnosis for Java services typically involves checking CPU, disk, memory, and network issues in sequence.

CPU problems are identified using ps, top -H -p PID, converting thread IDs to hex, and analyzing stack traces with jstack to locate high‑CPU threads.

Frequent GC can be monitored with jstat -gc PID and GC logs; adjust heap sizes or GC parameters as needed.

Disk issues are examined using df -hl, iostat -d -k -x, and iotop to find heavy I/O processes, then tracing file descriptors via readlink and cat /proc/PID/io.

Memory analysis covers heap OOM, Metaspace OOM, and stack overflow, using jmap -histo:live PID, enabling -XX:+HeapDumpOnOutOfMemoryError, and analyzing dumps with MAT or jcmd native memory tracking.

Off‑heap memory leaks are tracked with pmap -x PID, gdb dumps, and NMT via -XX:NativeMemoryTracking=summary.

Network problems such as timeouts, TCP queue overflow, RST packets, TIME_WAIT and CLOSE_WAIT states are diagnosed with netstat, ss, tcpdump, and kernel parameters like net.ipv4.tcp_tw_reuse and net.ipv4.tcp_tw_recycle.