Diagnosing and Resolving an Nginx Memory Leak on Linux Using Common Tools

A colleague reported that an Nginx instance on a server was consuming 4 GB of RAM, prompting an immediate investigation.

Step 1: Run top -u work to verify overall memory usage and confirm the 4 GB consumption.

Step 2: Use ps -ef | grep nginx | grep -v grep | grep work to locate the specific Nginx worker processes and identify which ones are using excessive memory.

Step 3: Examine error.log to understand why only certain workers show large memory footprints, indicating that the issue is not uniform across all workers.

Step 4: Check system logs with dmesg | grep pid to see that low‑memory workers are being OOM‑killed and the master process is respawning new workers.

Step 5: After confirming the leaking process, inspect its memory allocation:

Dump detailed allocation with pmap -x <pid>.

View memory segment boundaries via cat /proc/<pid>/smaps.

Attach with gdb -p <pid> and dump binary memory, e.g.,

dump binary memory ./memory2.log 0x7fa1d0b57000 0x7FA1D0B70000

.

Step 6: Analyze the dumped data and discover that a custom Nginx extension storing business data was responsible for the leak; fixing the extension and redeploying resolves the issue.

Common Linux process memory analysis commands: top – shows overall and per‑process memory usage (RES, SHR, VIRT, DATA). pmap -x <pid> – displays detailed memory map of a process. mtrace – traces memory allocation calls. gdb -p <pid> – attaches to a running process for debugging. cat /proc/<pid>/smaps – reveals start/end addresses of memory regions.

dump binary memory ./out.log <start_addr> <end_addr>

– extracts raw memory contents.