pre-delete pluging: update to naming conventions, add more tests

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!23

slapos/grid/SlapObject.py

@@ -354,6 +354,7 @@ class Partition(object):
     self.instance_path = instance_path
     self.run_path = os.path.join(self.instance_path, 'etc', 'run')
     self.service_path = os.path.join(self.instance_path, 'etc', 'service')
+    self.prerm_path = os.path.join(self.instance_path, 'etc', 'prerm')
     self.supervisord_partition_configuration_path = \
         supervisord_partition_configuration_path
     self.supervisord_socket = supervisord_socket
@@ -453,16 +454,18 @@ class Partition(object):
               'USER': pwd.getpwuid(uid).pw_name,
           }
 
-  def addServiceToCustomGroup(self, group_id, runner_list, path):
+  def addServiceToCustomGroup(self, group_suffix, partition_id, runner_list,
+      path, extension=''):
     """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 % {
+    group_id = '-'.join([partition_id, group_suffix])
+    self.supervisor_configuration_group += 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)
+    return self.addServiceToGroup(group_id, runner_list, path, extension)
 
   def updateSymlink(self, sr_symlink, software_path):
     if os.path.lexists(sr_symlink):
@@ -612,7 +615,7 @@ class Partition(object):
     runner_list = []
     service_list = []
     self.partition_supervisor_configuration = ""
-    self.supervisor_configuration_groups = ""
+    self.supervisor_configuration_group = ""
     if os.path.exists(self.run_path):
       if os.path.isdir(self.run_path):
         runner_list = os.listdir(self.run_path)
@@ -628,7 +631,7 @@ 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.supervisor_configuration_groups = group_partition_template % {
+      self.supervisor_configuration_group = group_partition_template % {
           'instance_id': partition_id,
           'program_list': ','.join(['_'.join([partition_id, runner])
                                     for runner in runner_list + service_list])
@@ -642,10 +645,10 @@ class Partition(object):
     """
       Write supervisord configuration file and update supervisord
     """
-    if self.supervisor_configuration_groups and \
+    if self.supervisor_configuration_group and \
         self.partition_supervisor_configuration:
       updateFile(self.supervisord_partition_configuration_path,
-                 self.supervisor_configuration_groups +
+                 self.supervisor_configuration_group +
                  self.partition_supervisor_configuration)
     self.updateSupervisor()
 

slapos/grid/slapgrid.py

@@ -80,7 +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'
+COMPUTER_PARTITION_WAIT_LIST_FILENAME = '.slapos-report-wait-service-list'
 
 # XXX hardcoded watchdog_path
 WATCHDOG_PATH = '/opt/slapos/bin/slapos-watchdog'
@@ -1273,7 +1273,7 @@ stderr_logfile_backups=1
 
     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]
+        processes_list = [name.strip() for name in wait_f if name]
         # return True if one of process in the list is running
         return partition.checkProcessesFromStateList(processes_list,
                                                      state_list)

slapos/manager/predestroy.py → slapos/manager/prerm.py

@@ -9,7 +9,6 @@ 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."""
@@ -45,32 +44,31 @@ class Manager(object):
 
     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):
+    if not os.path.exists(partition.prerm_path):
       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_id = partition.partition_id
+    wrapper_list = [f for f in os.listdir(partition.prerm_path)
+                    if os.path.isfile(os.path.join(partition.prerm_path, f))]
+    if len(wrapper_list) > 0:
+      group_suffix = "prerm"
+      logger.info("Adding pre-delete scripts to supervisord...")
       partition.generateSupervisorConfiguration()
-      partition.addServiceToCustomGroup(group_name,
-                                        wipe_wrapper_list,
-                                        wipe_base_folder)
-  
+      partition.addServiceToCustomGroup(group_suffix,
+                                        partition_id,
+                                        wrapper_list,
+                                        partition.prerm_path)
       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'
+      for name in wrapper_list:
+        process_name = '-'.join([partition_id, group_suffix]) + ':' + name
+        process_list_string += '%s\n' % process_name
         status = supervisord.getProcessInfo(process_name)
         if status['start'] == 0:
           # process is not started yet
-          logger.info("Starting pre-destroy process %r..." % name)
+          logger.info("Starting pre-delete process %r..." % name)
           supervisord.startProcess(process_name, False)
 
       # ask to slapgrid to check theses scripts before destroy partition

slapos/tests/slapgrid.py

@@ -40,6 +40,7 @@ import time
 import unittest
 import urlparse
 import json
+import re
 
 import xml_marshaller
 from mock import patch
@@ -53,6 +54,7 @@ from slapos.grid import SlapObject
 from slapos.grid.SlapObject import WATCHDOG_MARK
 from slapos.slap.slap import COMPUTER_PARTITION_REQUEST_LIST_TEMPLATE_FILENAME
 import slapos.grid.SlapObject
+from slapos import manager as slapmanager
 
 import httmock
 
@@ -2415,69 +2417,214 @@ class TestSlapgridCPWithTransaction(MasterMixin, unittest.TestCase):
 
       self.assertFalse(os.path.exists(request_list_file))
 
-class TestSlapgridCPWithPreDeleteScript(MasterMixin, unittest.TestCase):
+class TestSlapgridReportWithPreDeleteScript(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
+  prerm_script_content = """#!/bin/sh
+
+echo "Running prerm script for this partition..."
+touch etc/prerm.txt
+for i in {1..2}
+do
+  echo "sleeping for 1s..."
+  sleep 1
+done
+echo "finished prerm script."
+rm etc/prerm.txt
+
+exit 0
+"""
+
+  def _wait_prerm_script_finished(self, base_path):
+    check_file = os.path.join(base_path, 'etc/prerm.txt')
+    limit = 10
+    count = 0
+    time.sleep(1)
+    while (count < limit) and os.path.exists(check_file):
+      time.sleep(1)
+      count += 1
+
+  def test_partition_destroy_with_pre_remove_service(self):
     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_dir = os.path.join(partition.partition_path, 'etc/prerm')
       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
-""")
+        f.write(self.prerm_script_content)
       os.chmod(pre_delete_script, 0754)
       self.assertInstanceDirectoryListEqual(['0'])
       self.assertItemsEqual(os.listdir(partition.partition_path),
-                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg', 'var',
+                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg',
                              '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'})
+      manager_list = slapmanager.from_config({'manager_list': 'prerm'})
       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)
+      # Assert partition directory is not destroyed (pre-delete is running)
       self.assertInstanceDirectoryListEqual(['0'])
       self.assertItemsEqual(os.listdir(partition.partition_path),
-                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg', 'var',
+                            ['.slapgrid', '.0_wrapper.log', 'buildout.cfg',
                              'etc', 'software_release', 'worked', '.slapos-retention-lock-delay',
-                             '.0-destroy_slapos_pre_delete.log', '.slapos-wait-services',
+                             '.0-prerm_slapos_pre_delete.log', '.slapos-report-wait-service-list',
                              '.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)
+      # wait until the pre-delete script is finished
+      self._wait_prerm_script_finished(partition.partition_path)
+
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is empty
+      self.assertInstanceDirectoryListEqual(['0'])
+      self.assertItemsEqual(os.listdir(partition.partition_path), [])
+
+  def test_partition_destroy_pre_remove_with_retention_lock(self):
+    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, 'etc/prerm')
+      pre_delete_script = os.path.join(pre_delete_dir, 'slapos_pre_delete')
+      partition.requested_state = 'started'
+      partition.filter_dict = {'retention_delay': 1.0 / (3600 * 24)}
+
+      self.assertEqual(self.grid.processComputerPartitionList(), slapgrid.SLAPGRID_SUCCESS)
+      self.assertTrue(os.path.exists(os.path.join(
+          partition.partition_path,
+          slapos.grid.SlapObject.Partition.retention_lock_delay_filename
+      )))
+
+      os.makedirs(pre_delete_dir, 0o700)
+      with open(pre_delete_script, 'w') as f:
+        f.write(self.prerm_script_content)
+      os.chmod(pre_delete_script, 0754)
+      self.assertTrue(os.path.exists(pre_delete_script))
+
+      manager_list = slapmanager.from_config({'manager_list': 'prerm'})
+      self.grid._manager_list = manager_list
+
+      partition.requested_state = 'destroyed'
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is not destroyed (retention-delay-lock)
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', 'buildout.cfg', 'etc', 'software_release',
+                             'worked', '.slapos-retention-lock-delay',
+                             '.slapos-retention-lock-date', '.slapos-request-transaction-0'])
+      self.assertTrue(os.path.exists(pre_delete_script))
+      self.assertTrue(os.path.exists(os.path.join(
+          partition.partition_path,
+          slapos.grid.SlapObject.Partition.retention_lock_date_filename
+      )))
+
+      time.sleep(1)
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+
+      # Assert partition directory is not destroyed (pre-delete is running)
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', 'buildout.cfg', 'etc', 'software_release',
+                             'worked', '.slapos-retention-lock-delay', '.slapos-retention-lock-date',
+                             '.0-prerm_slapos_pre_delete.log', '.slapos-report-wait-service-list',
+                             '.slapos-request-transaction-0'])
+
+      # wait until the pre-delete script is finished
+      self._wait_prerm_script_finished(partition.partition_path)
+
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is empty
+      self.assertItemsEqual(os.listdir(partition.partition_path), [])
+
+  def test_partition_destroy_pre_remove_script_not_stopped(self):
+    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, 'etc/prerm')
+      pre_delete_script = os.path.join(pre_delete_dir, 'slapos_pre_delete')
+      partition.requested_state = 'started'
+      self.assertEqual(self.grid.processComputerPartitionList(), slapgrid.SLAPGRID_SUCCESS)
+      os.makedirs(pre_delete_dir, 0o700)
+      with open(pre_delete_script, 'w') as f:
+        f.write(self.prerm_script_content)
+      os.chmod(pre_delete_script, 0754)
+      self.assertEqual(partition.state, 'started')
+      manager_list = slapmanager.from_config({'manager_list': 'prerm'})
+      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-delete is running)
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', 'buildout.cfg', 'etc', 'software_release',
+                             'worked', '.slapos-retention-lock-delay', '.slapos-request-transaction-0',
+                             '.0-prerm_slapos_pre_delete.log', '.slapos-report-wait-service-list'])
+
+      # wait until the pre-delete script is finished
+      self._wait_prerm_script_finished(partition.partition_path)
+      with open(os.path.join(partition.partition_path, '.0-prerm_slapos_pre_delete.log')) as f:
+        # the script is well finished...
+        self.assertTrue("finished prerm script." in f.read())
 
       self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
       # Assert partition directory is empty
       self.assertInstanceDirectoryListEqual(['0'])
       self.assertItemsEqual(os.listdir(partition.partition_path), [])
+
+  def test_partition_destroy_pre_remove_script_run_as_partition_user(self):
+    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, 'etc/prerm')
+      pre_delete_script = os.path.join(pre_delete_dir, 'slapos_pre_delete')
+      partition.requested_state = 'started'
+      self.assertEqual(self.grid.processComputerPartitionList(), slapgrid.SLAPGRID_SUCCESS)
+      os.makedirs(pre_delete_dir, 0o700)
+      with open(pre_delete_script, 'w') as f:
+        f.write(self.prerm_script_content)
+      os.chmod(pre_delete_script, 0754)
+
+      manager_list = slapmanager.from_config({'manager_list': 'prerm'})
+      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-delete is running)
+      self.assertItemsEqual(os.listdir(partition.partition_path),
+                            ['.slapgrid', 'buildout.cfg', 'etc', 'software_release',
+                             'worked', '.slapos-retention-lock-delay', '.slapos-request-transaction-0',
+                             '.0-prerm_slapos_pre_delete.log', '.slapos-report-wait-service-list'])
+
+      stat_info = os.stat(partition.partition_path)
+      uid = stat_info.st_uid
+      gid = stat_info.st_gid
+      supervisor_conf_file = os.path.join(self.instance_root,
+                                          'etc/supervisord.conf.d',
+                                          '%s.conf' % partition.name)
+      self.assertTrue(os.path.exists(supervisor_conf_file))
+      regex_user = r"user=(\d+)"
+      regex_group = r"group=(\d+)"
+      with open(supervisor_conf_file) as f:
+        config = f.read()
+        # search user uid in conf file
+        result = re.search(regex_user, config, re.DOTALL)
+        self.assertTrue(result is not None)
+        self.assertEqual(int(result.groups()[0]), uid)
+        # search user group gid in conf file
+        result = re.search(regex_group, config, re.DOTALL)
+        self.assertTrue(result is not None)
+        self.assertEqual(int(result.groups()[0]), gid)
+
+      # wait until the pre-delete script is finished
+      self._wait_prerm_script_finished(partition.partition_path)
+
+      self.assertEqual(self.grid.agregateAndSendUsage(), slapgrid.SLAPGRID_SUCCESS)
+      # Assert partition directory is empty
+      self.assertInstanceDirectoryListEqual(['0'])
+      self.assertItemsEqual(os.listdir(partition.partition_path), [])
+