How to determine why an application is slow if it is a running binary without logs?

If you have a running binary without logs and need to understand why the application is slow, you can use the following approaches:

1. Real-time profiling:

   • Use system monitoring tools (top, htop, pidstat) to assess CPU, memory, IO load.
   • Apply strace or dtrace to track system calls and identify blocking.

2. Taking a profile dump:

   • If the application is written in Go, you can send the SIGQUIT signal (kill -QUIT pid) to get goroutine stacks in stdout or stderr.
   • Use pprof if the application supports an HTTP profiling endpoint.

3. Analyzing thread states:

   • Use gdb or delve to attach to the process and view thread states, call stacks.

4. Checking for deadlocks and locks:

   • Goroutine stack analysis will show where goroutines are waiting (e.g., on mutexes, channels).

5. Resource monitoring:

   • Check if resources are exhausted: memory, file descriptors, network.

6. Temporary logging addition:

   • If possible, rebuild the binary with logging or add tracing points.

Thus, without logs, diagnostics boil down to analyzing the process state and system resources using external tools and signals.