Merge pull request #40614 from gtmanfred/tcp

add retries on authentications of the salt minion reconnecting

conf/minion

@@ -735,6 +735,10 @@
 #
 #zmq_monitor: False
 
+# Number of times to try to authenticate with the salt master when reconnecting
+# to the master
+#tcp_authentication_retries: 5
+
 ######      Module configuration      #####
 ###########################################
 # Salt allows for modules to be passed arbitrary configuration data, any data

doc/ref/configuration/minion.rst

@@ -2122,6 +2122,20 @@ ZeroMQ is installed.
 
 .. conf_minion:: failhard
 
+``tcp_authentication_retries``
+------------------------------
+
+Default: ``5``
+
+The number of times to retry authenticating with the salt master when it comes
+back online.
+
+Zeromq does a lot to make sure when connections come back online that they
+reauthenticate. The tcp transport should try to connect with a new connection
+if the old one times out on reauthenticating.
+
+`-1` for infinite tries.
+
 ``failhard``
 ------------
 

salt/config/__init__.py

@@ -945,6 +945,10 @@ VALID_OPTS = {
     # django auth
     'django_auth_path': str,
     'django_auth_settings': str,
+
+    # Number of times to try to auth with the master on a reconnect with the
+    # tcp transport
+    'tcp_authentication_retries': int,
 }
 
 # default configurations
@@ -1075,6 +1079,7 @@ DEFAULT_MINION_OPTS = {
     'file_buffer_size': 262144,
     'tcp_pub_port': 4510,
     'tcp_pull_port': 4511,
+    'tcp_authentication_retries': 5,
     'log_file': os.path.join(salt.syspaths.LOGS_DIR, 'minion'),
     'log_level': 'warning',
     'log_level_logfile': None,

salt/transport/tcp.py

@@ -408,7 +408,14 @@ class AsyncTCPPubChannel(salt.transport.mixins.auth.AESPubClientMixin, salt.tran
             raise tornado.gen.Return(True)
 
         if force_auth or not self.auth.authenticated:
-            yield self.auth.authenticate()
+            count = 0
+            while count <= self.opts['tcp_authentication_retries'] or self.opts['tcp_authentication_retries'] < 0:
+                try:
+                    yield self.auth.authenticate()
+                    break
+                except SaltClientError as exc:
+                    log.debug(exc)
+                    count += 1
         try:
             ret = yield _do_transfer()
             raise tornado.gen.Return(ret)