Connect with address (not host)

distribute_tasks.py

@@ -4,6 +4,7 @@ import threading
 import queue
 import sys
 import time
+import socket
 import random
 import yaml
 import paramiko
@@ -58,8 +59,9 @@ class Task:
 class WorkingThread(threading.Thread):
     """ A thread actually getting work done on a given machine """
 
-    def __init__(self, host, workqueue, failures):
+    def __init__(self, host, addr, workqueue, failures):
         self.host = host
+        self.addr = addr
         self.client = None
         self.workqueue = workqueue
         self.failures = failures
@@ -70,18 +72,24 @@ class WorkingThread(threading.Thread):
         self.client.load_system_host_keys()
         for n_try in range(3):
             try:
-                self.client.connect(self.host, username=CONFIG["username"])
+                self.client.connect(self.addr, username=CONFIG["username"])
                 break
             except Exception as exn:
                 delay = 3 + random.random() * 4
                 print(
                     (
-                        "[{}] Failed to connect. Retry in {} seconds."
+                        "[{}] Failed to connect. Retry in {:.02f} seconds."
                         + "Exception:\n{}"
                     ).format(self.host, delay, exn),
                     file=sys.stderr,
                 )
                 time.sleep(delay)
+        else:
+            print(
+                "[{}] Failed to connect, stopping thread.".format(self.host),
+                file=sys.stderr,
+            )
+            return
 
         try:
             while True:
@@ -121,7 +129,10 @@ class HostsFile:
                             "Host {} has no {}".format(entry["host"], field)
                         )
                     raise Exception("Host has no {}".format(field))
-            self.hosts[entry["host"]] = {"cores": entry["cores"]}
+            self.hosts[entry["host"]] = {
+                "cores": entry["cores"],
+                "ip": socket.gethostbyname(entry["host"]),
+            }
 
 
 class TasksFile: