AMBARI-18733. Perf: Allow running multiple Ambari Agents on the same host (alejandro)

ambari-agent/conf/unix/ambari-agent

@@ -33,9 +33,17 @@ case "${1:-}" in
         ;;
 esac
 
+export HOME_DIR=""
+if [ "$#" = 3 ] && [ $2 = "--home" ] ; then
+  export HOME_DIR=$3
+  echo "Allow running multiple agents on this host; will use custom Home Dir: $HOME_DIR"
+fi
+
+export CONFIG_FILE="$HOME_DIR/etc/ambari-agent/conf/ambari-agent.ini"
+
 get_agent_property() {
   property_name="$1"
-  value=$(awk -F "=" "/^$property_name/ {print \$2}" /etc/ambari-agent/conf/ambari-agent.ini)
+  value=$(awk -F "=" "/^$property_name/ {print \$2}" $CONFIG_FILE)
   echo $value
 }
 
@@ -51,7 +59,7 @@ valid_path() {
 
 
 export PATH=/usr/sbin:/sbin:/usr/lib/ambari-server/*:$PATH
-export AMBARI_CONF_DIR=/etc/ambari-server/conf:$PATH
+export AMBARI_CONF_DIR=$HOME_DIR/etc/ambari-server/conf:$PATH
 
 # Because Ambari rpm unpacks modules here on all systems
 export PYTHONPATH=/usr/lib/python2.6/site-packages:${PYTHONPATH:-}

ambari-agent/conf/windows/service_wrapper.py

@@ -124,6 +124,7 @@ def ctrlHandler(ctrlType):
 #
 def setup(options):
   config = AmbariConfig()
+  # TODO AMBARI-18733, need to read home_dir to get correct config file.
   configFile = config.getConfigFile()
 
   updateConfigServerHostname(configFile, options.host_name)

ambari-agent/src/main/python/ambari_agent/AmbariConfig.py

@@ -188,7 +188,7 @@ class AmbariConfig:
 
   @staticmethod
   @OsFamilyFuncImpl(OSConst.WINSRV_FAMILY)
-  def getConfigFile():
+  def getConfigFile(home_dir=""):
     if 'AMBARI_AGENT_CONF_DIR' in os.environ:
       return os.path.join(os.environ['AMBARI_AGENT_CONF_DIR'], "ambari-agent.ini")
     else:
@@ -196,32 +196,56 @@ class AmbariConfig:
 
   @staticmethod
   @OsFamilyFuncImpl(OsFamilyImpl.DEFAULT)
-  def getConfigFile():
+  def getConfigFile(home_dir=""):
+    """
+    Get the configuration file path.
+    :param home_dir: In production, will be "". When running multiple Agents per host, each agent will have a unique path.
+    :return: Configuration file path.
+    """
     if 'AMBARI_AGENT_CONF_DIR' in os.environ:
       return os.path.join(os.environ['AMBARI_AGENT_CONF_DIR'], "ambari-agent.ini")
     else:
-      return os.path.join(os.sep, "etc", "ambari-agent", "conf", "ambari-agent.ini")
+      # home_dir may be an empty string
+      return os.path.join(os.sep, home_dir, "etc", "ambari-agent", "conf", "ambari-agent.ini")
 
+  # TODO AMBARI-18733, change usages of this function to provide the home_dir.
   @staticmethod
-  def getLogFile():
+  def getLogFile(home_dir=""):
+    """
+    Get the log file path.
+    :param home_dir: In production, will be "". When running multiple Agents per host, each agent will have a unique path.
+    :return: Log file path.
+    """
     if 'AMBARI_AGENT_LOG_DIR' in os.environ:
       return os.path.join(os.environ['AMBARI_AGENT_LOG_DIR'], "ambari-agent.log")
     else:
-      return os.path.join(os.sep, "var", "log", "ambari-agent", "ambari-agent.log")
-    
+      return os.path.join(os.sep, home_dir, "var", "log", "ambari-agent", "ambari-agent.log")
+
+  # TODO AMBARI-18733, change usages of this function to provide the home_dir.
   @staticmethod
-  def getAlertsLogFile():
+  def getAlertsLogFile(home_dir=""):
+    """
+    Get the alerts log file path.
+    :param home_dir: In production, will be "". When running multiple Agents per host, each agent will have a unique path.
+    :return: Alerts log file path.
+    """
     if 'AMBARI_AGENT_LOG_DIR' in os.environ:
       return os.path.join(os.environ['AMBARI_AGENT_LOG_DIR'], "ambari-agent.log")
     else:
-      return os.path.join(os.sep, "var", "log", "ambari-agent", "ambari-alerts.log")
+      return os.path.join(os.sep, home_dir, "var", "log", "ambari-agent", "ambari-alerts.log")
 
+  # TODO AMBARI-18733, change usages of this function to provide the home_dir.
   @staticmethod
-  def getOutFile():
+  def getOutFile(home_dir=""):
+    """
+    Get the out file path.
+    :param home_dir: In production, will be "". When running multiple Agents per host, each agent will have a unique path.
+    :return: Out file path.
+    """
     if 'AMBARI_AGENT_LOG_DIR' in os.environ:
       return os.path.join(os.environ['AMBARI_AGENT_LOG_DIR'], "ambari-agent.out")
     else:
-      return os.path.join(os.sep, "var", "log", "ambari-agent", "ambari-agent.out")
+      return os.path.join(os.sep, home_dir, "var", "log", "ambari-agent", "ambari-agent.out")
 
   def has_option(self, section, option):
     return self.config.has_option(section, option)

ambari-agent/src/main/python/ambari_agent/Facter.py

@@ -53,10 +53,18 @@ def run_os_command(cmd):
 
 
 class Facter(object):
-  def __init__(self):
-    self.config = self.resolve_ambari_config()
-  
+  def __init__(self, config):
+    """
+    Initialize the configs, which can be provided if using multiple Agents per host.
+    :param config: Agent configs. None if will use the default location.
+    """
+    self.config = config if config is not None else self.resolve_ambari_config()
+
   def resolve_ambari_config(self):
+    """
+    Resolve the default Ambari Agent configs.
+    :return: The default configs.
+    """
     try:
       config = AmbariConfig()
       if os.path.exists(AmbariConfig.getConfigFile()):
@@ -370,8 +378,8 @@ class FacterLinux(Facter):
   GET_UPTIME_CMD = "cat /proc/uptime"
   GET_MEMINFO_CMD = "cat /proc/meminfo"
 
-  def __init__(self):
-    super(FacterLinux,self).__init__()
+  def __init__(self, config):
+    super(FacterLinux,self).__init__(config)
     self.DATA_IFCONFIG_SHORT_OUTPUT = FacterLinux.setDataIfConfigShortOutput()
     self.DATA_UPTIME_OUTPUT = FacterLinux.setDataUpTimeOutput()
     self.DATA_MEMINFO_OUTPUT = FacterLinux.setMemInfoOutput()
@@ -547,7 +555,8 @@ class FacterLinux(Facter):
 
 
 def main(argv=None):
-  print Facter().facterInfo()
+  config = None
+  print Facter(config).facterInfo()
 
 
 if __name__ == '__main__':

ambari-agent/src/main/python/ambari_agent/Hardware.py

@@ -42,11 +42,12 @@ class Hardware:
   IGNORE_ROOT_MOUNTS = ["proc", "dev", "sys"]
   IGNORE_DEVICES = ["proc", "tmpfs", "cgroup", "mqueue", "shm"]
 
-  def __init__(self):
+  def __init__(self, config):
     self.hardware = {
       'mounts': Hardware.osdisks()
     }
-    self.hardware.update(Facter().facterInfo())
+    self.config = config
+    self.hardware.update(Facter(self.config).facterInfo())
 
   @classmethod
   def _parse_df_line(cls, line):
@@ -169,7 +170,8 @@ def main():
   from resource_management.core.logger import Logger
   Logger.initialize_logger()
 
-  print Hardware().get()
+  config = None
+  print Hardware(config).get()
 
 if __name__ == '__main__':
   main()

ambari-agent/src/main/python/ambari_agent/Heartbeat.py

@@ -111,8 +111,9 @@ def main(argv=None):
   from ambari_agent.Controller import Controller
 
   cfg = AmbariConfig()
-  if os.path.exists(AmbariConfig.getConfigFile()):
-    cfg.read(AmbariConfig.getConfigFile())
+  config_file_path = AmbariConfig.getConfigFile(home_dir="")
+  if os.path.exists(config_file_path):
+    cfg.read(config_file_path)
   else:
     raise Exception("No config found, use default")
 

ambari-agent/src/main/python/ambari_agent/Register.py

@@ -31,8 +31,8 @@ class Register:
   """ Registering with the server. Get the hardware profile and
   declare success for now """
   def __init__(self, config):
-    self.hardware = Hardware()
     self.config = config
+    self.hardware = Hardware(self.config)
 
   def build(self, version, id='-1'):
     global clusterId, clusterDefinitionRevision, firstContact

ambari-agent/src/main/python/ambari_agent/main.py

@@ -71,7 +71,12 @@ alerts_logger = logging.getLogger('ambari_alerts')
 
 formatstr = "%(levelname)s %(asctime)s %(filename)s:%(lineno)d - %(message)s"
 agentPid = os.getpid()
+
+# Global variables to be set later.
+home_dir = ""
+
 config = AmbariConfig.AmbariConfig()
+# TODO AMBARI-18733, remove this global variable and calculate it based on home_dir once it is set.
 configFile = config.getConfigFile()
 two_way_ssl_property = config.TWO_WAY_SSL_PROPERTY
 
@@ -109,7 +114,8 @@ def add_syslog_handler(logger):
 def update_log_level(config):
   # Setting loglevel based on config file
   global logger
-  log_cfg_file = os.path.join(os.path.dirname(AmbariConfig.AmbariConfig.getConfigFile()), "logging.conf")
+  global home_dir
+  log_cfg_file = os.path.join(os.path.dirname(AmbariConfig.AmbariConfig.getConfigFile(home_dir)), "logging.conf")
   if os.path.exists(log_cfg_file):
     logging.config.fileConfig(log_cfg_file)
     # create logger
@@ -131,11 +137,15 @@ def update_log_level(config):
       logger.info("Default loglevel=DEBUG")
 
 
-#  ToDo: move that function inside AmbariConfig
+# TODO AMBARI-18733, move inside AmbariConfig
 def resolve_ambari_config():
+  """
+  Load the configurations.
+  In production, home_dir will be "". When running multiple Agents per host, each agent will have a unique path.
+  """
   global config
-  configPath = os.path.abspath(AmbariConfig.AmbariConfig.getConfigFile())
-
+  global home_dir
+  configPath = os.path.abspath(AmbariConfig.AmbariConfig.getConfigFile(home_dir))
   try:
     if os.path.exists(configPath):
       config.read(configPath)
@@ -236,9 +246,11 @@ def stop_agent():
     sys.exit(0)
 
 def reset_agent(options):
+  global home_dir
   try:
     # update agent config file
     agent_config = ConfigParser.ConfigParser()
+    # TODO AMBARI-18733, calculate configFile based on home_dir
     agent_config.read(configFile)
     server_host = agent_config.get('server', 'hostname')
     new_host = options[2]
@@ -284,13 +296,17 @@ def run_threads(server_hostname, heartbeat_stop_callback):
 # we need this for windows os, where no sigterm available
 def main(heartbeat_stop_callback=None):
   global config
+  global home_dir
+
   parser = OptionParser()
   parser.add_option("-v", "--verbose", dest="verbose", action="store_true", help="verbose log output", default=False)
   parser.add_option("-e", "--expected-hostname", dest="expected_hostname", action="store",
                     help="expected hostname of current host. If hostname differs, agent will fail", default=None)
+  parser.add_option("--home", dest="home_dir", action="store", help="Home directory", default="")
   (options, args) = parser.parse_args()
 
   expected_hostname = options.expected_hostname
+  home_dir = options.home_dir
 
   logging_level = logging.DEBUG if options.verbose else logging.INFO
 
@@ -300,6 +316,10 @@ def main(heartbeat_stop_callback=None):
   setup_logging(alerts_logger, AmbariConfig.AmbariConfig.getAlertsLogFile(), logging_level)
   Logger.initialize_logger('resource_management', logging_level=logging_level)
 
+  if home_dir != "":
+    # When running multiple Ambari Agents on this host for simulation, each one will use a unique home directory.
+    Logger.info("Agent is using Home Dir: %s" % str(home_dir))
+
   # use the host's locale for numeric formatting
   try:
     locale.setlocale(locale.LC_ALL, '')

ambari-agent/src/test/python/ambari_agent/TestCheckWebUI.py

@@ -24,6 +24,8 @@ import sys
 
 from mock.mock import MagicMock, patch
 
+# Needed to import checkWebUI.py
+sys.path.append("../../../../ambari-server/src/main/resources/common-services/HDFS/2.1.0.2.0/package/files")
 import checkWebUI
 
 class TestMain(unittest.TestCase):

ambari-agent/src/test/python/ambari_agent/TestHardware.py

@@ -52,7 +52,8 @@ class TestHardware(TestCase):
   def test_build(self, get_os_version_mock, get_os_type_mock):
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    hardware = Hardware()
+    config = None
+    hardware = Hardware(config)
     result = hardware.get()
     osdisks = hardware.osdisks()
     for dev_item in result['mounts']:
@@ -196,7 +197,8 @@ class TestHardware(TestCase):
     hostname_mock.return_value = 'ambari'
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
 
     self.assertEquals(result['hostname'], "ambari")
     self.assertEquals(result['domain'], "apache.org")
@@ -211,7 +213,8 @@ class TestHardware(TestCase):
     facter_setDataUpTimeOutput_mock.return_value = "262813.00 123.45"
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
 
     self.assertEquals(result['uptime_seconds'], '262813')
     self.assertEquals(result['uptime_hours'], '73')
@@ -236,7 +239,8 @@ SwapFree:        1598676 kB
 
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
 
     self.assertEquals(result['memorysize'], 1832392)
     self.assertEquals(result['memorytotal'], 1832392)
@@ -261,7 +265,8 @@ SwapFree:        1598676 kB
 
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
 
     self.assertTrue(inet_ntoa_mock.called)
     self.assertTrue(get_ip_address_by_ifname_mock.called)
@@ -287,7 +292,8 @@ SwapFree:        1598676 kB
 
     get_os_type_mock.return_value = "suse"
     get_os_version_mock.return_value = "11"
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
 
     self.assertTrue(get_ip_address_by_ifname_mock.called)
     self.assertEquals(result['netmask'], None)
@@ -301,22 +307,23 @@ SwapFree:        1598676 kB
     get_os_version_mock.return_value = "11"
     get_os_family_mock.return_value = "redhat"
 
-    result = Facter().facterInfo()
+    config = None
+    result = Facter(config).facterInfo()
     self.assertEquals(result['operatingsystem'], 'some_type_of_os')
     self.assertEquals(result['osfamily'], 'redhat')
 
     get_os_family_mock.return_value = "ubuntu"
-    result = Facter().facterInfo()
+    result = Facter(config).facterInfo()
     self.assertEquals(result['operatingsystem'], 'some_type_of_os')
     self.assertEquals(result['osfamily'], 'ubuntu')
 
     get_os_family_mock.return_value = "suse"
-    result = Facter().facterInfo()
+    result = Facter(config).facterInfo()
     self.assertEquals(result['operatingsystem'], 'some_type_of_os')
     self.assertEquals(result['osfamily'], 'suse')
 
     get_os_family_mock.return_value = "My_new_family"
-    result = Facter().facterInfo()
+    result = Facter(config).facterInfo()
     self.assertEquals(result['operatingsystem'], 'some_type_of_os')
     self.assertEquals(result['osfamily'], 'My_new_family')
 
@@ -352,8 +359,7 @@ SwapFree:        1598676 kB
     json_data.items.return_value = [('key', 'value')]
     json_data.__getitem__.return_value = 'value'
 
-    facter = Facter()
-    facter.config = config
+    facter = Facter(config)
     result = facter.getSystemResourceOverrides()
 
     isdir.assert_called_with('/etc/custom_resource_overrides')

ambari-agent/src/test/python/ambari_agent/TestMain.py

@@ -95,10 +95,12 @@ class TestMain(unittest.TestCase):
     setLevel_mock.assert_called_with(logging.DEBUG)
 
 
+  @patch("os.path.exists")
   @patch.object(OSCheck, "os_distribution", new = MagicMock(return_value = os_distro_value))
   @patch.object(main.logger, "setLevel")
   @patch("logging.basicConfig")
-  def test_update_log_level(self, basicConfig_mock, setLevel_mock):
+  def test_update_log_level(self, basicConfig_mock, setLevel_mock, os_path_exists_mock):
+    os_path_exists_mock.return_value = False
     config = AmbariConfig().getConfig()
 
     # Testing with default setup (config file does not contain loglevel entry)