jax-ml/jax · error · ValueError

No active profiler server.

Error message

No active profiler server.

What it means

jax.profiler.stop_server raises ValueError when no profiler server is currently running; the global `_profiler_server` is None. It is the counterpart guard to start_server's singleton check.

Source

Thrown at jax/_src/profiler.py:89

  # Make sure backends are initialized before creating a profiler
  # session. Otherwise on Cloud TPU, libtpu may not be initialized before
  # creating the tracer, which will cause the TPU tracer initialization to
  # fail and no TPU operations will be included in the profile.
  # NOTE(skyewm): I'm not sure this is necessary for start_server (is definitely
  # is for start_trace), but I'm putting it here to be safe.
  if requires_backend:
    xla_bridge.get_backend()

  _profiler_server = _profiler.start_server(port)
  return _profiler_server


def stop_server():
  """Stops the running profiler server."""
  global _profiler_server
  if _profiler_server is None:
    raise ValueError("No active profiler server.")
  _profiler_server = None # Should destroy the profiler server


def register_subprocess(pid: int, port: int) -> Callable[[], None]:
  """Registers a subprocess's profiler server to be profiled alongside the current process.

  When the current process collects a profile (either programmatically or via
  its profiler server), it will propagate the request to all registered
  subprocesses' profiler servers and subsequently, aggregate all their responses
  into the main response returned by this process's profiler server.

  This is helpful when running workers in separate processes that may affect the
  performance of the main process (e.g. PyGrain).

  NOTE: Currently, only CPU profiling of subprocesses is supported.

  Args:
    pid: The process ID of the subprocess.

View on GitHub (pinned to 1e1c6a8fc0)

Solutions

1. Track server state in your own flag and only stop when started
2. Wrap stop_server in try/except ValueError for idempotent cleanup
3. Ensure tearDown only runs after a successful setUp start

Example fix

# before
jax.profiler.stop_server()  # may raise
# after
try:
  jax.profiler.stop_server()
except ValueError:
  pass

import jax.profiler as jp
# track your own state
server_started = False

try:
    jp.stop_server()
except ValueError:
    pass  # nothing to stop; idempotent cleanup

When it happens

Trigger: Calling `jax.profiler.stop_server()` before any successful start_server, or after stop_server was already called.

Common situations: Cleanup code (finally blocks, tearDown) running when startup failed or was skipped; paired tests where ordering varies.

Understand the failure class

Background: "Invalid state transition" errors: "status must be X, actually Y", "already rejected/charging/uninstalled", "cannot ... while running" — what they mean when a library rejects your call — this error's family across 31 libraries.

Related errors

• No profile started
• distributed.initialize should only be called once.
• Only one profiler server can be active at a time.
• Profile has already been started. Only one profile may be ru
• Invalid trace folder: {latest_trace_folder}

AI-assisted analysis of jax-ml/jax@1e1c6a8fc0 (2026-08-27). Data as JSON: /api/errors/ca2e4dba632b4093. Report an issue: GitHub.