Improve debugging when waiting for servers

Add debug messages and timers for start and end waiting for servers

and improve logic for awaiting proxy endpoints using concurrency primitives instead of a for-loop

jupyterhub/proxy.py

@@ -48,7 +48,7 @@ from jupyterhub.traitlets import Command
 from . import utils
 from .metrics import CHECK_ROUTES_DURATION_SECONDS, PROXY_POLL_DURATION_SECONDS
 from .objects import Server
-from .utils import AnyTimeoutError, exponential_backoff, url_escape_path, url_path_join
+from .utils import exponential_backoff, url_escape_path, url_path_join
 
 
 def _one_at_a_time(method):
@@ -766,24 +766,36 @@ class ConfigurableHTTPProxy(Proxy):
 
         self._write_pid_file()
 
-        def _check_process():
+        async def watch_process():
+            """Watch proxy process for early termination
+
+            so we don't keep waiting for connections after the proxy exits
+            """
+            while True:
                 status = self.proxy_process.poll()
                 if status is not None:
                     with self.proxy_process:
-                    e = RuntimeError("Proxy failed to start with exit code %i" % status)
+                        e = RuntimeError(
+                            f"Proxy failed to start with exit code {status}"
+                        )
                         raise e from None
 
-        for server in (public_server, api_server):
+                await asyncio.wait([servers_ready], timeout=0.5)
-            for i in range(10):
+                # stop watching the process when the connections
-                _check_process()
+                # have finished (success or failure)
-                try:
+                if servers_ready.done():
-                    await server.wait_up(1)
+                    return
-                except AnyTimeoutError:
+
-                    continue
+        servers_ready = asyncio.gather(
-                else:
+            *(server.wait_up(10) for server in (public_server, api_server))
-                    break
+        )
-            await server.wait_up(1)
+
-        _check_process()
+        process_ended = asyncio.ensure_future(watch_process())
+        # wait for process to crash or servers to be ready,
+        # whichever comes first
+        # don't need to handle exceptions here,
+        # as the underlying exceptions have informative error messages
+        await asyncio.wait_for(asyncio.gather(process_ended, servers_ready), timeout=15)
         self.log.debug("Proxy started and appears to be up")
         pc = PeriodicCallback(self.check_running, 1e3 * self.check_running_interval)
         self._check_running_callback = pc

jupyterhub/utils.py

@@ -15,6 +15,7 @@ import ssl
 import string
 import sys
 import threading
+import time
 import uuid
 import warnings
 from binascii import b2a_hex
@@ -93,6 +94,8 @@ def can_connect(ip, port):
     except OSError as e:
         if e.errno not in {errno.ECONNREFUSED, errno.ETIMEDOUT}:
             app_log.error("Unexpected error connecting to %s:%i %s", ip, port, e)
+        else:
+            app_log.debug("Server at %s:%i not ready: %s", ip, port, e)
         return False
     else:
         return True
@@ -245,6 +248,8 @@ async def wait_for_server(ip, port, timeout=10):
     """Wait for any server to show up at ip:port."""
     if ip in {'', '0.0.0.0', '::'}:
         ip = '127.0.0.1'
+    app_log.debug("Waiting %ss for server at %s:%s", timeout, ip, port)
+    tic = time.perf_counter()
     await exponential_backoff(
         lambda: can_connect(ip, port),
         "Server at {ip}:{port} didn't respond in {timeout} seconds".format(
@@ -252,6 +257,8 @@ async def wait_for_server(ip, port, timeout=10):
         ),
         timeout=timeout,
     )
+    toc = time.perf_counter()
+    app_log.debug("Server at %s:%s responded in %.2fs", ip, port, toc - tic)
 
 
 async def wait_for_http_server(url, timeout=10, ssl_context=None):
@@ -259,12 +266,13 @@ async def wait_for_http_server(url, timeout=10, ssl_context=None):
 
     Any non-5XX response code will do, even 404.
     """
-    loop = ioloop.IOLoop.current()
-    tic = loop.time()
     client = AsyncHTTPClient()
     if ssl_context:
         client.ssl_options = ssl_context
 
+    app_log.debug("Waiting %ss for server at %s", timeout, url)
+    tic = time.perf_counter()
+
     async def is_reachable():
         try:
             r = await client.fetch(url, follow_redirects=False)
@@ -297,6 +305,8 @@ async def wait_for_http_server(url, timeout=10, ssl_context=None):
         ),
         timeout=timeout,
     )
+    toc = time.perf_counter()
+    app_log.debug("Server at %s responded in %.2fs", url, toc - tic)
     return re