Synnefo » snf-ganeti

#

#

# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.

#

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

#

# This program is distributed in the hope that it will be useful, but

# WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

# General Public License for more details.

#

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA

# 02110-1301, USA.

"""Tool to restart erroneously downed virtual machines.

This program and set of classes implement a watchdog to restart

virtual machines in a Ganeti cluster that have crashed or been killed

by a node reboot.  Run from cron or similar.

"""

import os

import os.path

import sys

import time

import logging

from optparse import OptionParser

from ganeti import utils

from ganeti import constants

from ganeti import compat

from ganeti import serializer

from ganeti import errors

from ganeti import opcodes

from ganeti import cli

from ganeti import luxi

from ganeti import rapi

from ganeti import netutils

import ganeti.rapi.client # pylint: disable-msg=W0611

import ganeti.watcher.nodemaint # pylint: disable-msg=W0611

MAXTRIES = 5

# Delete any record that is older than 8 hours; this value is based on

# the fact that the current retry counter is 5, and watcher runs every

# 5 minutes, so it takes around half an hour to exceed the retry

# counter, so 8 hours (16*1/2h) seems like a reasonable reset time

RETRY_EXPIRATION = 8 * 3600

BAD_STATES = [constants.INSTST_ERRORDOWN]

HELPLESS_STATES = [constants.INSTST_NODEDOWN, constants.INSTST_NODEOFFLINE]

NOTICE = 'NOTICE'

ERROR = 'ERROR'

KEY_RESTART_COUNT = "restart_count"

KEY_RESTART_WHEN = "restart_when"

KEY_BOOT_ID = "bootid"

# Global LUXI client object

client = None

class NotMasterError(errors.GenericError):

  """Exception raised when this host is not the master."""

def ShouldPause():

  """Check whether we should pause.

  """

  return bool(utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE))

def StartNodeDaemons():

  """Start all the daemons that should be running on all nodes.

  """

  # on master or not, try to start the node daemon

  utils.EnsureDaemon(constants.NODED)

  # start confd as well. On non candidates it will be in disabled mode.

  utils.EnsureDaemon(constants.CONFD)

def RunWatcherHooks():

  """Run the watcher hooks.

  """

  hooks_dir = utils.PathJoin(constants.HOOKS_BASE_DIR,

                             constants.HOOKS_NAME_WATCHER)

  if not os.path.isdir(hooks_dir):

    return

  try:

    results = utils.RunParts(hooks_dir)

  except Exception: # pylint: disable-msg=W0703

    logging.exception("RunParts %s failed: %s", hooks_dir)

    return

  for (relname, status, runresult) in results:

    if status == constants.RUNPARTS_SKIP:

      logging.debug("Watcher hook %s: skipped", relname)

    elif status == constants.RUNPARTS_ERR:

      logging.warning("Watcher hook %s: error (%s)", relname, runresult)

    elif status == constants.RUNPARTS_RUN:

      if runresult.failed:

        logging.warning("Watcher hook %s: failed (exit: %d) (output: %s)",

                        relname, runresult.exit_code, runresult.output)

      else:

        logging.debug("Watcher hook %s: success (output: %s)", relname,

                      runresult.output)

class WatcherState(object):

  """Interface to a state file recording restart attempts.

  """

  def __init__(self, statefile):

    """Open, lock, read and parse the file.

    @type statefile: file

    @param statefile: State file object

    """

    self.statefile = statefile

    try:

      state_data = self.statefile.read()

      if not state_data:

        self._data = {}

      else:

        self._data = serializer.Load(state_data)

    except Exception, msg: # pylint: disable-msg=W0703

      # Ignore errors while loading the file and treat it as empty

      self._data = {}

      logging.warning(("Invalid state file. Using defaults."

                       " Error message: %s"), msg)

    if "instance" not in self._data:

      self._data["instance"] = {}

    if "node" not in self._data:

      self._data["node"] = {}

    self._orig_data = serializer.Dump(self._data)

  def Save(self):

    """Save state to file, then unlock and close it.

    """

    assert self.statefile

    serialized_form = serializer.Dump(self._data)

    if self._orig_data == serialized_form:

      logging.debug("Data didn't change, just touching status file")

      os.utime(constants.WATCHER_STATEFILE, None)

      return

    # We need to make sure the file is locked before renaming it, otherwise

    # starting ganeti-watcher again at the same time will create a conflict.

    fd = utils.WriteFile(constants.WATCHER_STATEFILE,

                         data=serialized_form,

                         prewrite=utils.LockFile, close=False)

    self.statefile = os.fdopen(fd, 'w+')

  def Close(self):

    """Unlock configuration file and close it.

    """

    assert self.statefile

    # Files are automatically unlocked when closing them

    self.statefile.close()

    self.statefile = None

  def GetNodeBootID(self, name):

    """Returns the last boot ID of a node or None.

    """

    ndata = self._data["node"]

    if name in ndata and KEY_BOOT_ID in ndata[name]:

      return ndata[name][KEY_BOOT_ID]

    return None

  def SetNodeBootID(self, name, bootid):

    """Sets the boot ID of a node.

    """

    assert bootid

    ndata = self._data["node"]

    if name not in ndata:

      ndata[name] = {}

    ndata[name][KEY_BOOT_ID] = bootid

  def NumberOfRestartAttempts(self, instance):

    """Returns number of previous restart attempts.

    @type instance: L{Instance}

    @param instance: the instance to look up

    """

    idata = self._data["instance"]

    if instance.name in idata:

      return idata[instance.name][KEY_RESTART_COUNT]

    return 0

  def MaintainInstanceList(self, instances):

    """Perform maintenance on the recorded instances.

    @type instances: list of string

    @param instances: the list of currently existing instances

    """

    idict = self._data["instance"]

    # First, delete obsolete instances

    obsolete_instances = set(idict).difference(instances)

    for inst in obsolete_instances:

      logging.debug("Forgetting obsolete instance %s", inst)

      del idict[inst]

    # Second, delete expired records

    earliest = time.time() - RETRY_EXPIRATION

    expired_instances = [i for i in idict

                         if idict[i][KEY_RESTART_WHEN] < earliest]

    for inst in expired_instances:

      logging.debug("Expiring record for instance %s", inst)

      del idict[inst]

  def RecordRestartAttempt(self, instance):

    """Record a restart attempt.

    @type instance: L{Instance}

    @param instance: the instance being restarted

    """

    idata = self._data["instance"]

    if instance.name not in idata:

      inst = idata[instance.name] = {}

    else:

      inst = idata[instance.name]

    inst[KEY_RESTART_WHEN] = time.time()

    inst[KEY_RESTART_COUNT] = inst.get(KEY_RESTART_COUNT, 0) + 1

  def RemoveInstance(self, instance):

    """Update state to reflect that a machine is running.

    This method removes the record for a named instance (as we only

    track down instances).

    @type instance: L{Instance}

    @param instance: the instance to remove from books

    """

    idata = self._data["instance"]

    if instance.name in idata:

      del idata[instance.name]

class Instance(object):

  """Abstraction for a Virtual Machine instance.

  """

  def __init__(self, name, state, autostart, snodes):

    self.name = name

    self.state = state

    self.autostart = autostart

    self.snodes = snodes

  def Restart(self):

    """Encapsulates the start of an instance.

    """

    op = opcodes.OpInstanceStartup(instance_name=self.name, force=False)

    cli.SubmitOpCode(op, cl=client)

  def ActivateDisks(self):

    """Encapsulates the activation of all disks of an instance.

    """

    op = opcodes.OpInstanceActivateDisks(instance_name=self.name)

    cli.SubmitOpCode(op, cl=client)

def GetClusterData():

  """Get a list of instances on this cluster.

  """

  op1_fields = ["name", "status", "admin_state", "snodes"]

  op1 = opcodes.OpInstanceQuery(output_fields=op1_fields, names=[],

                                use_locking=True)

  op2_fields = ["name", "bootid", "offline"]

  op2 = opcodes.OpNodeQuery(output_fields=op2_fields, names=[],

                            use_locking=True)

  job_id = client.SubmitJob([op1, op2])

  all_results = cli.PollJob(job_id, cl=client, feedback_fn=logging.debug)

  logging.debug("Got data from cluster, writing instance status file")

  result = all_results[0]

  smap = {}

  instances = {}

  # write the instance status file

  up_data = "".join(["%s %s\n" % (fields[0], fields[1]) for fields in result])

  utils.WriteFile(file_name=constants.INSTANCE_STATUS_FILE, data=up_data)

  for fields in result:

    (name, status, autostart, snodes) = fields

    # update the secondary node map

    for node in snodes:

      if node not in smap:

        smap[node] = []

      smap[node].append(name)

    instances[name] = Instance(name, status, autostart, snodes)

  nodes =  dict([(name, (bootid, offline))

                 for name, bootid, offline in all_results[1]])

  client.ArchiveJob(job_id)

  return instances, nodes, smap

class Watcher(object):

  """Encapsulate the logic for restarting erroneously halted virtual machines.

  The calling program should periodically instantiate me and call Run().

  This will traverse the list of instances, and make up to MAXTRIES attempts

  to restart machines that are down.

  """

  def __init__(self, opts, notepad):

    self.notepad = notepad

    master = client.QueryConfigValues(["master_node"])[0]

    if master != netutils.Hostname.GetSysName():

      raise NotMasterError("This is not the master node")

    # first archive old jobs

    self.ArchiveJobs(opts.job_age)

    # and only then submit new ones

    self.instances, self.bootids, self.smap = GetClusterData()

    self.started_instances = set()

    self.opts = opts

  def Run(self):

    """Watcher run sequence.

    """

    notepad = self.notepad

    self.CheckInstances(notepad)

    self.CheckDisks(notepad)

    self.VerifyDisks()

  @staticmethod

  def ArchiveJobs(age):

    """Archive old jobs.

    """

    arch_count, left_count = client.AutoArchiveJobs(age)

    logging.debug("Archived %s jobs, left %s", arch_count, left_count)

  def CheckDisks(self, notepad):

    """Check all nodes for restarted ones.

    """

    check_nodes = []

    for name, (new_id, offline) in self.bootids.iteritems():

      old = notepad.GetNodeBootID(name)

      if new_id is None:

        # Bad node, not returning a boot id

        if not offline:

          logging.debug("Node %s missing boot id, skipping secondary checks",

                        name)

        continue

      if old != new_id:

        # Node's boot ID has changed, proably through a reboot.

        check_nodes.append(name)

    if check_nodes:

      # Activate disks for all instances with any of the checked nodes as a

      # secondary node.

      for node in check_nodes:

        if node not in self.smap:

          continue

        for instance_name in self.smap[node]:

          instance = self.instances[instance_name]

          if not instance.autostart:

            logging.info(("Skipping disk activation for non-autostart"

                          " instance %s"), instance.name)

            continue

          if instance.name in self.started_instances:

            # we already tried to start the instance, which should have

            # activated its drives (if they can be at all)

            logging.debug("Skipping disk activation for instance %s, as"

                          " it was already started", instance.name)

            continue

          try:

            logging.info("Activating disks for instance %s", instance.name)

            instance.ActivateDisks()

          except Exception: # pylint: disable-msg=W0703

            logging.exception("Error while activating disks for instance %s",

                              instance.name)

      # Keep changed boot IDs

      for name in check_nodes:

        notepad.SetNodeBootID(name, self.bootids[name][0])

  def CheckInstances(self, notepad):

    """Make a pass over the list of instances, restarting downed ones.

    """

    notepad.MaintainInstanceList(self.instances.keys())

    for instance in self.instances.values():

      if instance.state in BAD_STATES:

        n = notepad.NumberOfRestartAttempts(instance)

        if n > MAXTRIES:

          logging.warning("Not restarting instance %s, retries exhausted",

                          instance.name)

          continue

        elif n < MAXTRIES:

          last = " (Attempt #%d)" % (n + 1)

        else:

          notepad.RecordRestartAttempt(instance)

          logging.error("Could not restart %s after %d attempts, giving up",

                        instance.name, MAXTRIES)

          continue

        try:

          logging.info("Restarting %s%s", instance.name, last)

          instance.Restart()

          self.started_instances.add(instance.name)

        except Exception: # pylint: disable-msg=W0703

          logging.exception("Error while restarting instance %s",

                            instance.name)

        notepad.RecordRestartAttempt(instance)

      elif instance.state in HELPLESS_STATES:

        if notepad.NumberOfRestartAttempts(instance):

          notepad.RemoveInstance(instance)

      else:

        if notepad.NumberOfRestartAttempts(instance):

          notepad.RemoveInstance(instance)

          logging.info("Restart of %s succeeded", instance.name)

  def _CheckForOfflineNodes(self, instance):

    """Checks if given instances has any secondary in offline status.

    @param instance: The instance object

    @return: True if any of the secondary is offline, False otherwise

    """

    bootids = []

    for node in instance.snodes:

      bootids.append(self.bootids[node])

    return compat.any(offline for (_, offline) in bootids)

  def VerifyDisks(self):

    """Run gnt-cluster verify-disks.

    """

    job_id = client.SubmitJob([opcodes.OpClusterVerifyDisks()])

    result = cli.PollJob(job_id, cl=client, feedback_fn=logging.debug)[0]

    client.ArchiveJob(job_id)

    # Keep track of submitted jobs

    jex = cli.JobExecutor(cl=client, feedback_fn=logging.debug)

    archive_jobs = set()

    for (status, job_id) in result[constants.JOB_IDS_KEY]:

      jex.AddJobId(None, status, job_id)

      if status:

        archive_jobs.add(job_id)

    offline_disk_instances = set()

    for (status, result) in jex.GetResults():

      if not status:

        logging.error("Verify-disks job failed: %s", result)

        continue

      ((_, instances, _), ) = result

      offline_disk_instances.update(instances)

    for job_id in archive_jobs:

      client.ArchiveJob(job_id)

    if not offline_disk_instances:

      # nothing to do

      logging.debug("verify-disks reported no offline disks, nothing to do")

      return

    logging.debug("Will activate disks for instance(s) %s",

                  utils.CommaJoin(offline_disk_instances))

    # we submit only one job, and wait for it. not optimal, but spams

    # less the job queue

    job = []

    for name in offline_disk_instances:

      instance = self.instances[name]

      if (instance.state in HELPLESS_STATES or

          self._CheckForOfflineNodes(instance)):

        logging.info("Skip instance %s because it is in helpless state or has"

                     " one offline secondary", name)

        continue

      job.append(opcodes.OpInstanceActivateDisks(instance_name=name))

    if job:

      job_id = cli.SendJob(job, cl=client)

      try:

        cli.PollJob(job_id, cl=client, feedback_fn=logging.debug)

      except Exception: # pylint: disable-msg=W0703

        logging.exception("Error while activating disks")

def OpenStateFile(path):

  """Opens the state file and acquires a lock on it.

  @type path: string

  @param path: Path to state file

  """

  # The two-step dance below is necessary to allow both opening existing

  # file read/write and creating if not existing. Vanilla open will truncate

  # an existing file -or- allow creating if not existing.

  statefile_fd = os.open(path, os.O_RDWR | os.O_CREAT)

  # Try to acquire lock on state file. If this fails, another watcher instance

  # might already be running or another program is temporarily blocking the

  # watcher from running.

  try:

    utils.LockFile(statefile_fd)

  except errors.LockError, err:

    logging.error("Can't acquire lock on state file %s: %s", path, err)

    return None

  return os.fdopen(statefile_fd, "w+")

def IsRapiResponding(hostname):

  """Connects to RAPI port and does a simple test.

  Connects to RAPI port of hostname and does a simple test. At this time, the

  test is GetVersion.

  @type hostname: string

  @param hostname: hostname of the node to connect to.

  @rtype: bool

  @return: Whether RAPI is working properly

  """

  curl_config = rapi.client.GenericCurlConfig()

  rapi_client = rapi.client.GanetiRapiClient(hostname,

                                             curl_config_fn=curl_config)

  try:

    master_version = rapi_client.GetVersion()

  except rapi.client.CertificateError, err:

    logging.warning("RAPI Error: CertificateError (%s)", err)

    return False

  except rapi.client.GanetiApiError, err:

    logging.warning("RAPI Error: GanetiApiError (%s)", err)

    return False

  logging.debug("RAPI Result: master_version is %s", master_version)

  return master_version == constants.RAPI_VERSION

def ParseOptions():

  """Parse the command line options.

  @return: (options, args) as from OptionParser.parse_args()

  """

  parser = OptionParser(description="Ganeti cluster watcher",

                        usage="%prog [-d]",

                        version="%%prog (ganeti) %s" %

                        constants.RELEASE_VERSION)

  parser.add_option(cli.DEBUG_OPT)

  parser.add_option("-A", "--job-age", dest="job_age", default=6 * 3600,

                    help="Autoarchive jobs older than this age (default"

                          " 6 hours)")

  parser.add_option("--ignore-pause", dest="ignore_pause", default=False,

                    action="store_true", help="Ignore cluster pause setting")

  options, args = parser.parse_args()

  options.job_age = cli.ParseTimespec(options.job_age)

  if args:

    parser.error("No arguments expected")

  return (options, args)

@rapi.client.UsesRapiClient

def Main():

  """Main function.

  """

  global client # pylint: disable-msg=W0603

  (options, _) = ParseOptions()

  utils.SetupLogging(constants.LOG_WATCHER, sys.argv[0],

                     debug=options.debug, stderr_logging=options.debug)

  if ShouldPause() and not options.ignore_pause:

    logging.debug("Pause has been set, exiting")

    return constants.EXIT_SUCCESS

  statefile = OpenStateFile(constants.WATCHER_STATEFILE)

  if not statefile:

    return constants.EXIT_FAILURE

  update_file = False

  try:

    StartNodeDaemons()

    RunWatcherHooks()

    # run node maintenance in all cases, even if master, so that old

    # masters can be properly cleaned up too

    if nodemaint.NodeMaintenance.ShouldRun(): # pylint: disable-msg=E0602

      nodemaint.NodeMaintenance().Exec() # pylint: disable-msg=E0602

    notepad = WatcherState(statefile)

    try:

      try:

        client = cli.GetClient()

      except errors.OpPrereqError:

        # this is, from cli.GetClient, a not-master case

        logging.debug("Not on master, exiting")

        update_file = True

        return constants.EXIT_SUCCESS

      except luxi.NoMasterError, err:

        logging.warning("Master seems to be down (%s), trying to restart",

                        str(err))

        if not utils.EnsureDaemon(constants.MASTERD):

          logging.critical("Can't start the master, exiting")

          return constants.EXIT_FAILURE

        # else retry the connection

        client = cli.GetClient()

      # we are on master now

      utils.EnsureDaemon(constants.RAPI)

      # If RAPI isn't responding to queries, try one restart.

      logging.debug("Attempting to talk with RAPI.")

      if not IsRapiResponding(constants.IP4_ADDRESS_LOCALHOST):

        logging.warning("Couldn't get answer from Ganeti RAPI daemon."

                        " Restarting Ganeti RAPI.")

        utils.StopDaemon(constants.RAPI)

        utils.EnsureDaemon(constants.RAPI)

        logging.debug("Second attempt to talk with RAPI")

        if not IsRapiResponding(constants.IP4_ADDRESS_LOCALHOST):

          logging.fatal("RAPI is not responding. Please investigate.")

      logging.debug("Successfully talked to RAPI.")

      try:

        watcher = Watcher(options, notepad)

      except errors.ConfigurationError:

        # Just exit if there's no configuration

        update_file = True

        return constants.EXIT_SUCCESS

      watcher.Run()

      update_file = True

    finally:

      if update_file:

        notepad.Save()

      else:

        logging.debug("Not updating status file due to failure")

  except SystemExit:

    raise

  except NotMasterError:

    logging.debug("Not master, exiting")

    return constants.EXIT_NOTMASTER

  except errors.ResolverError, err:

    logging.error("Cannot resolve hostname '%s', exiting.", err.args[0])

    return constants.EXIT_NODESETUP_ERROR

  except errors.JobQueueFull:

    logging.error("Job queue is full, can't query cluster state")

  except errors.JobQueueDrainError:

    logging.error("Job queue is drained, can't maintain cluster state")

  except Exception, err:

    logging.exception(str(err))

    return constants.EXIT_FAILURE

  return constants.EXIT_SUCCESS