WIP: erp5-util supervisord from previous run

erp5/tests/testERP5TestNode.py

 import unittest
+import mock
 from unittest import TestCase
 from contextlib import contextmanager
 
@@ -1138,3 +1139,29 @@ shared = true
     RunnerClass.updateDictionaryFile = original_updateDictionaryFile
     RunnerClass._createInstance = original__createInstance
     RunnerClass._waitInstanceCreation = original__waitInstanceCreation
+
+  def test_cleanup_supervisord_from_previous_run(self):
+    test_node = self.getTestNode()
+    runner = test_type_registry['UnitTest'](test_node)
+    node_test_suite = test_node.getNodeTestSuite('foo')
+    host, port = '127.0.0.1', 1234 # XXX
+    test_node.config['ipv4_address'] = host
+    test_node.config['ipv6_address'] = '::1' # TODO
+    test_node.config['proxy_host'] = host
+    test_node.config['proxy_port'] = port
+    test_node.config['master_url'] = "http://{proxy_host}:{proxy_port}".format(**test_node.config)
+    test_node.config['slapos_binary'] = '/usr/bin/slapos' # XXX must be in path (and not be the slaprunner wrapper !)
+    test_node.config['partition_reference'] = self.id()
+    test_node.config['environment'] = os.environ
+    test_node.config["software_list"] = []
+
+    with mock.patch(
+          'erp5.util.testnode.UnitTestRunner.SlapOSControler.runSoftwareRelease',
+          return_value={'status_code': 0}): #,\
+     #   mock.patch(
+     #     'erp5.util.testnode.UnitTestRunner.SlapOSControler.runComputerPartitionXXX',
+     #     return_value={'status_code': 0}):
+      runner.prepareSlapOSForTestSuite(node_test_suite)
+      import pdb; pdb.set_trace()
+      test_node.process_manager.max_timeout = 1
+      test_node.process_manager.spawn("sleep" "10")
\ No newline at end of file

erp5/util/testnode/ProcessManager.py

@@ -142,6 +142,7 @@ class ProcessManager(object):
     self.timer_set = set()
 
   def spawn(self, *args, **kw):
+    print "spawn", args, self.max_timeout
     def timeoutExpired(p):
       if p.poll() is None:
         logger.warning('PROCESS TOO LONG OR DEAD, GOING TO BE TERMINATED')

erp5/util/testnode/SlapOSControler.py

@@ -40,13 +40,14 @@ MAX_SR_RETRIES = 3
 
 class SlapOSControler(object):
 
-  def __init__(self, working_directory, config):
+  def __init__(self, working_directory, config, logger=None):
     self.config = config
     self.software_root = os.path.join(working_directory, 'soft')
     self.instance_root = os.path.join(working_directory, 'inst')
     self.slapos_config = os.path.join(working_directory, 'slapos.cfg')
     self.proxy_database = os.path.join(working_directory, 'proxy.db')
     self.instance_config = {}
+    self.logger = logger
 
   #TODO: implement a method to get all instance related the slapOS account
   # and deleting all old instances (based on creation date or name etc...)
@@ -54,7 +55,7 @@ class SlapOSControler(object):
   def createSlaposConfigurationFileAccount(self, key, certificate, slapos_url, config):
     # Create "slapos_account" directory in the "slapos_directory"
     slapos_account_directory = os.path.join(config['slapos_directory'], "slapos_account")
-    createFolder(slapos_account_directory)
+    createFolder(slapos_account_directory, logger)
     # Create slapos-account files
     slapos_account_key_path = os.path.join(slapos_account_directory, "key")
     slapos_account_certificate_path = os.path.join(slapos_account_directory, "certificate")
@@ -204,7 +205,7 @@ class SlapOSControler(object):
   def _resetSoftware(self):
     logger.info('SlapOSControler: GOING TO RESET ALL SOFTWARE : %r',
              self.software_root)
-    createFolder(self.software_root, True)
+    createFolder(self.software_root, True, logger)
 
   def initializeSlapOSControler(self, slapproxy_log=None, process_manager=None,
         reset_software=False, software_path_list=None):
@@ -266,12 +267,12 @@ class SlapOSControler(object):
     if reset_software:
       self._resetSoftware()
     else:
-      createFolder(self.software_root)
+      createFolder(self.software_root, logger)
     instance_root = self.instance_root
     # Delete any existing partition in order to not get its data (ex.
     # MySQL DB content) from previous runs. To support changes of partition
     # naming scheme (which already happened), do this at instance_root level.
-    createFolder(instance_root, True)
+    createFolder(instance_root, True, logger)
     for i in xrange(MAX_PARTITIONS):
       # create partition and configure computer
       # XXX: at the moment all partitions do share same virtual interface address
@@ -308,9 +309,10 @@ class SlapOSControler(object):
     os.environ['PATH'] = environment['PATH']
     # a SR may fail for number of reasons (incl. network failures)
     # so be tolerant and run it a few times before giving up
+    #import pdb; pdb.set_trace()
     for _ in xrange(MAX_SR_RETRIES):
       status_dict = self.spawn(config['slapos_binary'],
-                 'node', 'software', '--all', 
+                 'node', 'software',  # '--all',  XXX all does not uses .completed
                  '--pidfile', os.path.join(self.software_root, 'slapos-node.pid'),
                  '--cfg', self.slapos_config, raise_error_if_fail=False,
                  log_prefix='slapgrid_sr', get_output=False)

erp5/util/testnode/Utils.py

 import os
 import stat
 import shutil
+import psutil
+import logging
 
 def rmtree(path):
   """Delete a path recursively.
@@ -26,10 +28,25 @@ def rmtree(path):
   shutil.rmtree(path, onerror=chmod_retry)
 
 
-def createFolder(folder, clean=False):
+
+def createFolder(folder, clean=False, logger=None):
   if os.path.exists(folder):
     if not clean:
       return
+
+    # before removing, check that we don't have a supervisord left here,
+    # and terminate it if it's the case
+    supervisord_pid_file = os.path.join(folder, 'var/run/supervisord.pid')
+    if os.path.exists(supervisord_pid_file):
+      with open(supervisord_pid_file, 'r') as f:
+        pid = int(f.read().strip())
+        supervisord = psutil.Process(pid)
+        if supervisord.cmdline()[-1].endswith("supervisor.supervisord.main()"):
+          if logger is None:
+            logger = logging.getLogger()
+          logger.warn("Killing previous supervisord with pid %s" % pid)
+          supervisord.terminate()
+    
     rmtree(folder)
   os.mkdir(folder)