add pluging for run instance script at partition pre-destroy phase

This pluging can be used a add a script which wipe data before the partition is removed
The script run after all partition services are stopped. The script is ran by supervisord
then slapos node report will wait until it finish before destroy the partition.

/reviewed-on nexedi/slapos.core!22

slapos/grid/SlapObject.py

@@ -453,6 +453,17 @@ class Partition(object):
               'USER': pwd.getpwuid(uid).pw_name,
           }
 
+  def addServiceToCustomGroup(self, group_id, runner_list, path):
+    """Add new services to supervisord that belong to specific group"""
+    group_partition_template = pkg_resources.resource_stream(__name__,
+      'templates/group_partition_supervisord.conf.in').read()
+    self.supervisor_configuration_groups += group_partition_template % {
+        'instance_id': group_id,
+        'program_list': ','.join(['_'.join([group_id, runner])
+                                  for runner in runner_list])
+    }
+    return self.addServiceToGroup(group_id, runner_list, path)
+
   def updateSymlink(self, sr_symlink, software_path):
     if os.path.lexists(sr_symlink):
       if not os.path.islink(sr_symlink):
@@ -589,7 +600,7 @@ class Partition(object):
     self.generateSupervisorConfigurationFile()
     self.createRetentionLockDelay()
 
-  def generateSupervisorConfigurationFile(self):
+  def generateSupervisorConfiguration(self):
     """
     Generates supervisord configuration file from template.
 
@@ -600,6 +611,8 @@ class Partition(object):
     """
     runner_list = []
     service_list = []
+    self.partition_supervisor_configuration = ""
+    self.supervisor_configuration_groups = ""
     if os.path.exists(self.run_path):
       if os.path.isdir(self.run_path):
         runner_list = os.listdir(self.run_path)
@@ -615,7 +628,7 @@ class Partition(object):
       partition_id = self.computer_partition.getId()
       group_partition_template = pkg_resources.resource_stream(__name__,
           'templates/group_partition_supervisord.conf.in').read()
-      self.partition_supervisor_configuration = group_partition_template % {
+      self.supervisor_configuration_groups = group_partition_template % {
           'instance_id': partition_id,
           'program_list': ','.join(['_'.join([partition_id, runner])
                                     for runner in runner_list + service_list])
@@ -624,10 +637,25 @@ class Partition(object):
       self.addServiceToGroup(partition_id, runner_list, self.run_path)
       self.addServiceToGroup(partition_id, service_list, self.service_path,
                              extension=WATCHDOG_MARK)
+
+  def writeSupervisorConfigurationFile(self):
+    """
+      Write supervisord configuration file and update supervisord
+    """
+    if self.supervisor_configuration_groups and \
+        self.partition_supervisor_configuration:
       updateFile(self.supervisord_partition_configuration_path,
+                 self.supervisor_configuration_groups +
                  self.partition_supervisor_configuration)
     self.updateSupervisor()
 
+  def generateSupervisorConfigurationFile(self):
+    """
+      update supervisord with new processes
+    """
+    self.generateSupervisorConfiguration()
+    self.writeSupervisorConfigurationFile()
+
   def start(self):
     """Asks supervisord to start the instance. If this instance is not
     installed, we install it.
@@ -720,6 +748,19 @@ class Partition(object):
 
     return True
 
+  def checkProcessesFromStateList(self, process_list, state_list):
+    """Asks supervisord to check if one of the processes are in the state_list."""
+    supervisor = self.getSupervisorRPC()
+    for process in process_list:
+      try:
+        info = supervisor.getProcessInfo(process)
+        if info['statename'] in state_list:
+          return True
+      except xmlrpclib.Fault as exc:
+        self.logger.debug("BAD process name: %r" % process)
+        continue
+    return False
+
   def cleanupFolder(self, folder_path):
     """Delete all files and folders in a specified directory
     """

slapos/grid/slapgrid.py

@@ -80,6 +80,7 @@ PROMISE_TIMEOUT = 3
 COMPUTER_PARTITION_TIMESTAMP_FILENAME = '.timestamp'
 COMPUTER_PARTITION_LATEST_BANG_TIMESTAMP_FILENAME = '.slapos_latest_bang_timestamp'
 COMPUTER_PARTITION_INSTALL_ERROR_FILENAME = '.slapgrid-%s-error.log'
+COMPUTER_PARTITION_WAIT_LIST_FILENAME = '.slapos-wait-services'
 
 # XXX hardcoded watchdog_path
 WATCHDOG_PATH = '/opt/slapos/bin/slapos-watchdog'
@@ -1266,6 +1267,18 @@ stderr_logfile_backups=1
       return SLAPGRID_PROMISE_FAIL
     return SLAPGRID_SUCCESS
 
+  def _checkWaitProcessList(self, partition, state_list):
+    wait_file = os.path.join(partition.instance_path,
+                             COMPUTER_PARTITION_WAIT_LIST_FILENAME)
+
+    if os.path.exists(wait_file) and os.path.isfile(wait_file):
+      with open(wait_file) as wait_f:
+        processes_list = [name.strip() for name in wait_f.readlines() if name]
+        # return True if one of process in the list is running
+        return partition.checkProcessesFromStateList(processes_list,
+                                                     state_list)
+    return False
+
   def validateXML(self, to_be_validated, xsd_model):
     """Validates a given xml file"""
     #We retrieve the xsd model
@@ -1572,10 +1585,19 @@ stderr_logfile_backups=1
             raise
           except Exception:
             pass
+          # let managers update current partition
+          for manager in self._manager_list:
+            manager.report(local_partition)
+
           if computer_partition.getId() in report_usage_issue_cp_list:
             self.logger.info('Ignoring destruction of %r, as no report usage was sent' %
                                 computer_partition.getId())
             continue
+          if self._checkWaitProcessList(local_partition,
+              state_list=['RUNNING', 'STARTING']):
+            self.logger.info('There are running processes into the partition,' \
+              ' wait until they finish...')
+            continue
           destroyed = local_partition.destroy()
         except (SystemExit, KeyboardInterrupt):
           computer_partition.error(traceback.format_exc(), logger=self.logger)

slapos/manager/interface.py

@@ -29,3 +29,8 @@ class IManager(Interface):
     :param partition: slapos.grid.SlapObject.Partition, currently processed partition
     """
 
+  def report(partition):
+    """Method called at `slapos node report` phase.
+
+    :param partition: slapos.grid.SlapObject.Partition, currently processed partition
+    """

slapos/manager/predestroy.py

+# coding: utf-8
+import logging
+import os
+import sys
+import subprocess
+
+from zope import interface as zope_interface
+from slapos.manager import interface
+from slapos.grid.slapgrid import COMPUTER_PARTITION_WAIT_LIST_FILENAME
+
+logger = logging.getLogger(__name__)
+WIPE_WRAPPER_BASE_PATH = "var/run/slapos/pre-destroy/"
+
+class Manager(object):
+  """Manager is called in every step of preparation of the computer."""
+
+  zope_interface.implements(interface.IManager)
+
+  def __init__(self, config):
+    """Manager needs to know config for its functioning.
+    """
+    pass
+
+  def format(self, computer):
+    """Method called at `slapos node format` phase.
+    """
+    pass
+
+  def software(self, software):
+    """Method called at `slapos node software` phase.
+    """
+    pass
+
+  def instance(self, partition):
+    """Method called at `slapos node instance` phase.
+    """
+    pass
+
+  def report(self, partition):
+    """Method called at `slapos node report` phase."""
+
+    partition.createRetentionLockDate()
+    if not partition.checkRetentionIsAuthorized():
+      return
+
+    wait_filepath = os.path.join(partition.instance_path,
+                                 COMPUTER_PARTITION_WAIT_LIST_FILENAME)
+    wipe_base_folder = os.path.join(partition.instance_path,
+                                    WIPE_WRAPPER_BASE_PATH)
+    if not os.path.exists(wipe_base_folder):
+      return
+    wipe_wrapper_list = [f for f in os.listdir(wipe_base_folder)
+                         if os.path.isfile(os.path.join(wipe_base_folder, f))]
+    if len(wipe_wrapper_list) > 0:
+      group_name = partition.partition_id + '-' + "destroy"
+      logger.info("Adding pre-destroy scripts to supervisord...")
+      partition.generateSupervisorConfiguration()
+      partition.addServiceToCustomGroup(group_name,
+                                        wipe_wrapper_list,
+                                        wipe_base_folder)
+  
+      partition.writeSupervisorConfigurationFile()
+
+      # check the state of all process, if the process is not started yes, start it
+      supervisord = partition.getSupervisorRPC()
+      process_list_string = ""
+      for name in wipe_wrapper_list:
+        process_name = group_name + ':' + name
+        process_list_string += process_name + '\n'
+        status = supervisord.getProcessInfo(process_name)
+        if status['start'] == 0:
+          # process is not started yet
+          logger.info("Starting pre-destroy process %r..." % name)
+          supervisord.startProcess(process_name, False)
+
+      # ask to slapgrid to check theses scripts before destroy partition
+      with open(wait_filepath, 'w') as f:
+        f.write(process_list_string)

slapos/tests/slapgrid.py

@@ -2414,3 +2414,70 @@ class TestSlapgridCPWithTransaction(MasterMixin, unittest.TestCase):
       self.assertInstanceDirectoryListEqual(['0'])
 
       self.assertFalse(os.path.exists(request_list_file))
+
+class TestSlapgridCPWithPreDeleteScript(MasterMixin, unittest.TestCase):
+
+  def test_one_partition_pre_destroy_service(self):
+    from slapos import manager as slapmanager
+    from slapos.manager.predestroy import WIPE_WRAPPER_BASE_PATH
+    computer = ComputerForTest(self.software_root, self.instance_root)
+    with httmock.HTTMock(computer.request_handler):
+      partition = computer.instance_list[0]
+      pre_delete_dir = os.path.join(partition.partition_path, WIPE_WRAPPER_BASE_PATH)
+      pre_delete_script = os.path.join(pre_delete_dir, 'slapos_pre_delete')
+      partition.requested_state = 'started'
+      partition.software.setBuildout(WRAPPER_CONTENT)
+      self.assertEqual(self.grid.processComputerPartitionList(), slapgrid.SLAPGRID_SUCCESS)
+      os.makedirs(pre_delete_dir, 0o700)
+      with open(pre_delete_script, 'w') as f:
+        f.write("""#!/bin/sh
+
+echo "Running script to wipe this partition..."
+
+for i in {1..3}
+do
+  echo "sleeping for 1s..."
+  sleep 1
+done
+
+echo "finished wipe disk."
+
+exit 0
+""")
+      os.chmod(pre_delete_script, 0754)
+      self.assertInstanceDirectoryListEqual(['0'])
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg', 'var',
+                             'etc', 'software_release', 'worked', '.slapos-retention-lock-delay'])
+      wrapper_log = os.path.join(partition.partition_path, '.0_wrapper.log')
+      self.assertLogContent(wrapper_log, 'Working')
+      self.assertItemsEqual(os.listdir(self.software_root), [partition.software.software_hash])
+      self.assertEqual(computer.sequence,
+                       ['/getFullComputerInformation', '/availableComputerPartition',
+                        '/startedComputerPartition'])
+      self.assertEqual(partition.state, 'started')
+      partition.requested_state = 'stopped'
+      self.assertEqual(self.launchSlapgrid(), slapgrid.SLAPGRID_SUCCESS)
+      self.assertEqual(partition.state, 'stopped')
+      manager_list = slapmanager.from_config({'manager_list': 'predestroy'})
+      self.grid._manager_list = manager_list
+
+      partition.requested_state = 'destroyed'
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is not destroyed (pre-destroy is running)
+      self.assertInstanceDirectoryListEqual(['0'])
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg', 'var',
+                             'etc', 'software_release', 'worked', '.slapos-retention-lock-delay',
+                             '.0-destroy_slapos_pre_delete.log', '.slapos-wait-services',
+                             '.slapos-request-transaction-0'])
+      self.assertItemsEqual(os.listdir(self.software_root),
+                            [partition.software.software_hash])
+
+      # wait until the pre-destroy script is finished
+      time.sleep(5)
+
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is empty
+      self.assertInstanceDirectoryListEqual(['0'])
+      self.assertItemsEqual(os.listdir(partition.partition_path), [])