AMBARI-10819 Ambari bootstrap: Premature Agent shutdown may lead to Agent stall

Removed the agent stop invocation. The agent restarts now only after all the bits have been deployed.

ambari-agent/conf/unix/ambari-agent

@@ -220,8 +220,10 @@ case "$1" in
         ;;
   restart)
         echo -e "Restarting $AMBARI_AGENT"
-        $0 stop
-        $0 start "$@"
+        scriptpath=$0
+        $scriptpath stop
+        shift
+        $scriptpath start "$@"
         retcode=$?
         ;;
   reset)

ambari-server/src/main/python/setupAgent.py

@@ -182,21 +182,22 @@ def runAgent(passPhrase, expected_hostname, user_run_as, verbose, ret=None):
   return {"exitstatus": agent_retcode, "log": log}
 
 @OsFamilyFuncImpl(OSConst.WINSRV_FAMILY)
-def tryStopAgent():
+def checkVerbose():
   verbose = False
   if os.path.exists(AMBARI_AGENT_INSTALL_SYMLINK):
-    agentStopCmd = ["cmd", "/c", "ambari-agent.cmd", "stop"]
-    execOsCommand(agentStopCmd, tries=3, try_sleep=10, cwd=AMBARI_AGENT_INSTALL_SYMLINK)
+    agentStatusCmd = ["cmd", "/c", "ambari-agent.cmd", "status"]
+    ret = execOsCommand(agentStatusCmd, tries=3, try_sleep=10, cwd=AMBARI_AGENT_INSTALL_SYMLINK)
+    if ret["exitstatus"] == 0 and ret["log"][0].find("running") != -1:
+      verbose = True
   return verbose
 
 @OsFamilyFuncImpl(OsFamilyImpl.DEFAULT)
-def tryStopAgent():
+def checkVerbose():
   verbose = False
   cmds = ["bash", "-c", "ps aux | grep 'AmbariAgent.py' | grep ' \-v'"]
   cmdl = ["bash", "-c", "ps aux | grep 'AmbariAgent.py' | grep ' \--verbose'"]
   if execOsCommand(cmds)["exitstatus"] == 0 or execOsCommand(cmdl)["exitstatus"] == 0:
     verbose = True
-  subprocess.call("/usr/sbin/ambari-agent stop", shell=True)
   return verbose
 
 @OsFamilyFuncImpl(OSConst.WINSRV_FAMILY)
@@ -362,9 +363,9 @@ def run_setup(argv=None):
   return runAgent(passPhrase, expected_hostname, user_run_as, verbose)
 
 def main(argv=None):
-  #Try stop agent and check --verbose option if agent already run
+  #Check --verbose option if agent already running
   global verbose
-  verbose = tryStopAgent()
+  verbose = checkVerbose()
   if verbose:
     exitcode = run_setup(argv)
   else:

ambari-server/src/test/python/TestSetupAgent.py

@@ -252,7 +252,7 @@ class TestSetupAgent(TestCase):
     self.assertFalse(setup_agent.execOsCommand("hostname -f") == None)
 
   @patch.object(OSCheck, "os_distribution", new = MagicMock(return_value = os_distro_value))
-  @patch.object(setup_agent, 'tryStopAgent')
+  @patch.object(setup_agent, 'checkVerbose')
   @patch.object(setup_agent, 'isAgentPackageAlreadyInstalled')
   @patch.object(setup_agent, 'runAgent')
   @patch.object(setup_agent, 'configureAgent')
@@ -267,14 +267,14 @@ class TestSetupAgent(TestCase):
   def test_setup_agent_main(self, dirname_mock, realpath_mock, exit_mock, checkServerReachability_mock,
                             getOptimalVersion_mock, is_ubuntu_family_mock, is_suse_family_mock,
                             installAgent_mock, configureAgent_mock, runAgent_mock,
-                            isAgentPackageAlreadyInstalled_mock, tryStopAgent_mock):
+                            isAgentPackageAlreadyInstalled_mock, checkVerbose_mock):
     checkServerReachability_mock.return_value = {'log': 'log', 'exitstatus': 0}
     installAgent_mock.return_value = {'log': 'log', 'exitstatus': 0}
     configureAgent_mock.return_value = {'log': 'log', 'exitstatus': 0}
     runAgent_mock.return_value = {'log': 'log', 'exitstatus': 0}
     getOptimalVersion_mock.return_value = {'log': '1.1.2, 1.1.3, ', 'exitstatus': 1}
     ret = setup_agent.main(("setupAgent.py","agents_host","password", "server_hostname","1.1.1","8080"))
-    self.assertTrue(tryStopAgent_mock.called)
+    self.assertTrue(checkVerbose_mock.called)
     self.assertFalse(exit_mock.called)
     self.assertTrue("exitstatus" in ret)
     self.assertEqual(ret["exitstatus"], 1)