Synnefo » snf-ganeti

#

#

# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 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.

"""Inter-node RPC library.

"""

# pylint: disable=C0103,R0201,R0904

# C0103: Invalid name, since call_ are not valid

# R0201: Method could be a function, we keep all rpcs instance methods

# as not to change them back and forth between static/instance methods

# if they need to start using instance attributes

# R0904: Too many public methods

import logging

import zlib

import base64

import pycurl

import threading

import copy

import os

from ganeti import utils

from ganeti import objects

from ganeti import http

from ganeti import serializer

from ganeti import constants

from ganeti import errors

from ganeti import netutils

from ganeti import ssconf

from ganeti import runtime

from ganeti import compat

from ganeti import rpc_defs

from ganeti import pathutils

from ganeti import vcluster

# Special module generated at build time

from ganeti import _generated_rpc

# pylint has a bug here, doesn't see this import

import ganeti.http.client  # pylint: disable=W0611

_RPC_CLIENT_HEADERS = [

  "Content-type: %s" % http.HTTP_APP_JSON,

  "Expect:",

  ]

#: Special value to describe an offline host

_OFFLINE = object()

def Init():

  """Initializes the module-global HTTP client manager.

  Must be called before using any RPC function and while exactly one thread is

  running.

  """

  # curl_global_init(3) and curl_global_cleanup(3) must be called with only

  # one thread running. This check is just a safety measure -- it doesn't

  # cover all cases.

  assert threading.activeCount() == 1, \

         "Found more than one active thread when initializing pycURL"

  logging.info("Using PycURL %s", pycurl.version)

  pycurl.global_init(pycurl.GLOBAL_ALL)

def Shutdown():

  """Stops the module-global HTTP client manager.

  Must be called before quitting the program and while exactly one thread is

  running.

  """

  pycurl.global_cleanup()

def _ConfigRpcCurl(curl):

  noded_cert = str(pathutils.NODED_CERT_FILE)

  noded_client_cert = str(pathutils.NODED_CLIENT_CERT_FILE)

  # FIXME: The next two lines are necessary to ensure upgradability from

  # 2.10 to 2.11. Remove in 2.12, because this slows down RPC calls.

  if not os.path.exists(noded_client_cert):

    logging.info("Using server certificate as client certificate for RPC"

                 "call.")

    noded_client_cert = noded_cert

  curl.setopt(pycurl.FOLLOWLOCATION, False)

  curl.setopt(pycurl.CAINFO, noded_cert)

  curl.setopt(pycurl.SSL_VERIFYHOST, 0)

  curl.setopt(pycurl.SSL_VERIFYPEER, True)

  curl.setopt(pycurl.SSLCERTTYPE, "PEM")

  curl.setopt(pycurl.SSLCERT, noded_client_cert)

  curl.setopt(pycurl.SSLKEYTYPE, "PEM")

  curl.setopt(pycurl.SSLKEY, noded_client_cert)

  curl.setopt(pycurl.CONNECTTIMEOUT, constants.RPC_CONNECT_TIMEOUT)

def RunWithRPC(fn):

  """RPC-wrapper decorator.

  When applied to a function, it runs it with the RPC system

  initialized, and it shutsdown the system afterwards. This means the

  function must be called without RPC being initialized.

  """

  def wrapper(*args, **kwargs):

    Init()

    try:

      return fn(*args, **kwargs)

    finally:

      Shutdown()

  return wrapper

def _Compress(_, data):

  """Compresses a string for transport over RPC.

  Small amounts of data are not compressed.

  @type data: str

  @param data: Data

  @rtype: tuple

  @return: Encoded data to send

  """

  # Small amounts of data are not compressed

  if len(data) < 512:

    return (constants.RPC_ENCODING_NONE, data)

  # Compress with zlib and encode in base64

  return (constants.RPC_ENCODING_ZLIB_BASE64,

          base64.b64encode(zlib.compress(data, 3)))

class RpcResult(object):

  """RPC Result class.

  This class holds an RPC result. It is needed since in multi-node

  calls we can't raise an exception just because one out of many

  failed, and therefore we use this class to encapsulate the result.

  @ivar data: the data payload, for successful results, or None

  @ivar call: the name of the RPC call

  @ivar node: the name of the node to which we made the call

  @ivar offline: whether the operation failed because the node was

      offline, as opposed to actual failure; offline=True will always

      imply failed=True, in order to allow simpler checking if

      the user doesn't care about the exact failure mode

  @ivar fail_msg: the error message if the call failed

  """

  def __init__(self, data=None, failed=False, offline=False,

               call=None, node=None):

    self.offline = offline

    self.call = call

    self.node = node

    if offline:

      self.fail_msg = "Node is marked offline"

      self.data = self.payload = None

    elif failed:

      self.fail_msg = self._EnsureErr(data)

      self.data = self.payload = None

    else:

      self.data = data

      if not isinstance(self.data, (tuple, list)):

        self.fail_msg = ("RPC layer error: invalid result type (%s)" %

                         type(self.data))

        self.payload = None

      elif len(data) != 2:

        self.fail_msg = ("RPC layer error: invalid result length (%d), "

                         "expected 2" % len(self.data))

        self.payload = None

      elif not self.data[0]:

        self.fail_msg = self._EnsureErr(self.data[1])

        self.payload = None

      else:

        # finally success

        self.fail_msg = None

        self.payload = data[1]

    for attr_name in ["call", "data", "fail_msg",

                      "node", "offline", "payload"]:

      assert hasattr(self, attr_name), "Missing attribute %s" % attr_name

  def __repr__(self):

    return ("RpcResult(data=%s, call=%s, node=%s, offline=%s, fail_msg=%s)" %

            (self.offline, self.call, self.node, self.offline, self.fail_msg))

  @staticmethod

  def _EnsureErr(val):

    """Helper to ensure we return a 'True' value for error."""

    if val:

      return val

    else:

      return "No error information"

  def Raise(self, msg, prereq=False, ecode=None):

    """If the result has failed, raise an OpExecError.

    This is used so that LU code doesn't have to check for each

    result, but instead can call this function.

    """

    if not self.fail_msg:

      return

    if not msg: # one could pass None for default message

      msg = ("Call '%s' to node '%s' has failed: %s" %

             (self.call, self.node, self.fail_msg))

    else:

      msg = "%s: %s" % (msg, self.fail_msg)

    if prereq:

      ec = errors.OpPrereqError

    else:

      ec = errors.OpExecError

    if ecode is not None:

      args = (msg, ecode)

    else:

      args = (msg, )

    raise ec(*args) # pylint: disable=W0142

  def Warn(self, msg, feedback_fn):

    """If the result has failed, call the feedback_fn.

    This is used to in cases were LU wants to warn the

    user about a failure, but continue anyway.

    """

    if not self.fail_msg:

      return

    msg = "%s: %s" % (msg, self.fail_msg)

    feedback_fn(msg)

def _SsconfResolver(ssconf_ips, node_list, _,

                    ssc=ssconf.SimpleStore,

                    nslookup_fn=netutils.Hostname.GetIP):

  """Return addresses for given node names.

  @type ssconf_ips: bool

  @param ssconf_ips: Use the ssconf IPs

  @type node_list: list

  @param node_list: List of node names

  @type ssc: class

  @param ssc: SimpleStore class that is used to obtain node->ip mappings

  @type nslookup_fn: callable

  @param nslookup_fn: function use to do NS lookup

  @rtype: list of tuple; (string, string)

  @return: List of tuples containing node name and IP address

  """

  ss = ssc()

  family = ss.GetPrimaryIPFamily()

  if ssconf_ips:

    iplist = ss.GetNodePrimaryIPList()

    ipmap = dict(entry.split() for entry in iplist)

  else:

    ipmap = {}

  result = []

  for node in node_list:

    ip = ipmap.get(node)

    if ip is None:

      ip = nslookup_fn(node, family=family)

    result.append((node, ip, node))

  return result

class _StaticResolver:

  def __init__(self, addresses):

    """Initializes this class.

    """

    self._addresses = addresses

  def __call__(self, hosts, _):

    """Returns static addresses for hosts.

    """

    assert len(hosts) == len(self._addresses)

    return zip(hosts, self._addresses, hosts)

def _CheckConfigNode(node_uuid_or_name, node, accept_offline_node):

  """Checks if a node is online.

  @type node_uuid_or_name: string

  @param node_uuid_or_name: Node UUID

  @type node: L{objects.Node} or None

  @param node: Node object

  """

  if node is None:

    # Assume that the passed parameter was actually a node name, so depend on

    # DNS for name resolution

    return (node_uuid_or_name, node_uuid_or_name, node_uuid_or_name)

  else:

    if node.offline and not accept_offline_node:

      ip = _OFFLINE

    else:

      ip = node.primary_ip

    return (node.name, ip, node_uuid_or_name)

def _NodeConfigResolver(single_node_fn, all_nodes_fn, node_uuids, opts):

  """Calculate node addresses using configuration.

  Note that strings in node_uuids are treated as node names if the UUID is not

  found in the configuration.

  """

  accept_offline_node = (opts is rpc_defs.ACCEPT_OFFLINE_NODE)

  assert accept_offline_node or opts is None, "Unknown option"

  # Special case for single-host lookups

  if len(node_uuids) == 1:

    (uuid, ) = node_uuids

    return [_CheckConfigNode(uuid, single_node_fn(uuid), accept_offline_node)]

  else:

    all_nodes = all_nodes_fn()

    return [_CheckConfigNode(uuid, all_nodes.get(uuid, None),

                             accept_offline_node)

            for uuid in node_uuids]

class _RpcProcessor:

  def __init__(self, resolver, port, lock_monitor_cb=None):

    """Initializes this class.

    @param resolver: callable accepting a list of node UUIDs or hostnames,

      returning a list of tuples containing name, IP address and original name

      of the resolved node. IP address can be the name or the special value

      L{_OFFLINE} to mark offline machines.

    @type port: int

    @param port: TCP port

    @param lock_monitor_cb: Callable for registering with lock monitor

    """

    self._resolver = resolver

    self._port = port

    self._lock_monitor_cb = lock_monitor_cb

  @staticmethod

  def _PrepareRequests(hosts, port, procedure, body, read_timeout):

    """Prepares requests by sorting offline hosts into separate list.

    @type body: dict

    @param body: a dictionary with per-host body data

    """

    results = {}

    requests = {}

    assert isinstance(body, dict)

    assert len(body) == len(hosts)

    assert compat.all(isinstance(v, str) for v in body.values())

    assert frozenset(map(lambda x: x[2], hosts)) == frozenset(body.keys()), \

        "%s != %s" % (hosts, body.keys())

    for (name, ip, original_name) in hosts:

      if ip is _OFFLINE:

        # Node is marked as offline

        results[original_name] = RpcResult(node=name,

                                           offline=True,

                                           call=procedure)

      else:

        requests[original_name] = \

          http.client.HttpClientRequest(str(ip), port,

                                        http.HTTP_POST, str("/%s" % procedure),

                                        headers=_RPC_CLIENT_HEADERS,

                                        post_data=body[original_name],

                                        read_timeout=read_timeout,

                                        nicename="%s/%s" % (name, procedure),

                                        curl_config_fn=_ConfigRpcCurl)

    return (results, requests)

  @staticmethod

  def _CombineResults(results, requests, procedure):

    """Combines pre-computed results for offline hosts with actual call results.

    """

    for name, req in requests.items():

      if req.success and req.resp_status_code == http.HTTP_OK:

        host_result = RpcResult(data=serializer.LoadJson(req.resp_body),

                                node=name, call=procedure)

      else:

        # TODO: Better error reporting

        if req.error:

          msg = req.error

        else:

          msg = req.resp_body

        logging.error("RPC error in %s on node %s: %s", procedure, name, msg)

        host_result = RpcResult(data=msg, failed=True, node=name,

                                call=procedure)

      results[name] = host_result

    return results

  def __call__(self, nodes, procedure, body, read_timeout, resolver_opts,

               _req_process_fn=None):

    """Makes an RPC request to a number of nodes.

    @type nodes: sequence

    @param nodes: node UUIDs or Hostnames

    @type procedure: string

    @param procedure: Request path

    @type body: dictionary

    @param body: dictionary with request bodies per host

    @type read_timeout: int or None

    @param read_timeout: Read timeout for request

    @rtype: dictionary

    @return: a dictionary mapping host names to rpc.RpcResult objects

    """

    assert read_timeout is not None, \

      "Missing RPC read timeout for procedure '%s'" % procedure

    if _req_process_fn is None:

      _req_process_fn = http.client.ProcessRequests

    (results, requests) = \

      self._PrepareRequests(self._resolver(nodes, resolver_opts), self._port,

                            procedure, body, read_timeout)

    _req_process_fn(requests.values(), lock_monitor_cb=self._lock_monitor_cb)

    assert not frozenset(results).intersection(requests)

    return self._CombineResults(results, requests, procedure)

class _RpcClientBase:

  def __init__(self, resolver, encoder_fn, lock_monitor_cb=None,

               _req_process_fn=None):

    """Initializes this class.

    """

    proc = _RpcProcessor(resolver,

                         netutils.GetDaemonPort(constants.NODED),

                         lock_monitor_cb=lock_monitor_cb)

    self._proc = compat.partial(proc, _req_process_fn=_req_process_fn)

    self._encoder = compat.partial(self._EncodeArg, encoder_fn)

  @staticmethod

  def _EncodeArg(encoder_fn, node, (argkind, value)):

    """Encode argument.

    """

    if argkind is None:

      return value

    else:

      return encoder_fn(argkind)(node, value)

  def _Call(self, cdef, node_list, args):

    """Entry point for automatically generated RPC wrappers.

    """

    (procedure, _, resolver_opts, timeout, argdefs,

     prep_fn, postproc_fn, _) = cdef

    if callable(timeout):

      read_timeout = timeout(args)

    else:

      read_timeout = timeout

    if callable(resolver_opts):

      req_resolver_opts = resolver_opts(args)

    else:

      req_resolver_opts = resolver_opts

    if len(args) != len(argdefs):

      raise errors.ProgrammerError("Number of passed arguments doesn't match")

    if prep_fn is None:

      prep_fn = lambda _, args: args

    assert callable(prep_fn)

    # encode the arguments for each node individually, pass them and the node

    # name to the prep_fn, and serialise its return value

    encode_args_fn = lambda node: map(compat.partial(self._encoder, node),

                                      zip(map(compat.snd, argdefs), args))

    pnbody = dict(

      (n,

       serializer.DumpJson(prep_fn(n, encode_args_fn(n)),

                           private_encoder=serializer.EncodeWithPrivateFields))

      for n in node_list

    )

    result = self._proc(node_list, procedure, pnbody, read_timeout,

                        req_resolver_opts)

    if postproc_fn:

      return dict(map(lambda (key, value): (key, postproc_fn(value)),

                      result.items()))

    else:

      return result

def _ObjectToDict(_, value):

  """Converts an object to a dictionary.

  @note: See L{objects}.

  """

  return value.ToDict()

def _ObjectListToDict(node, value):

  """Converts a list of L{objects} to dictionaries.

  """

  return map(compat.partial(_ObjectToDict, node), value)

def _PrepareFileUpload(getents_fn, node, filename):

  """Loads a file and prepares it for an upload to nodes.

  """

  statcb = utils.FileStatHelper()

  data = _Compress(node, utils.ReadFile(filename, preread=statcb))

  st = statcb.st

  if getents_fn is None:

    getents_fn = runtime.GetEnts

  getents = getents_fn()

  virt_filename = vcluster.MakeVirtualPath(filename)

  return [virt_filename, data, st.st_mode, getents.LookupUid(st.st_uid),

          getents.LookupGid(st.st_gid), st.st_atime, st.st_mtime]

def _PrepareFinalizeExportDisks(_, snap_disks):

  """Encodes disks for finalizing export.

  """

  flat_disks = []

  for disk in snap_disks:

    if isinstance(disk, bool):

      flat_disks.append(disk)

    else:

      flat_disks.append(disk.ToDict())

  return flat_disks

def _EncodeBlockdevRename(_, value):

  """Encodes information for renaming block devices.

  """

  return [(d.ToDict(), uid) for d, uid in value]

def _AddSpindlesToLegacyNodeInfo(result, space_info):

  """Extracts the spindle information from the space info and adds

  it to the result dictionary.

  @type result: dict of strings

  @param result: dictionary holding the result of the legacy node info

  @type space_info: list of dicts of strings

  @param space_info: list, each row holding space information of one storage

    unit

  @rtype: None

  @return: does not return anything, manipulates the C{result} variable

  """

  lvm_pv_info = utils.storage.LookupSpaceInfoByStorageType(

      space_info, constants.ST_LVM_PV)

  if lvm_pv_info:

    result["spindles_free"] = lvm_pv_info["storage_free"]

    result["spindles_total"] = lvm_pv_info["storage_size"]

  else:

    result["spindles_free"] = 0

    result["spindles_total"] = 0

def _AddStorageInfoToLegacyNodeInfoByTemplate(

    result, space_info, disk_template):

  """Extracts the storage space information of the disk template from

  the space info and adds it to the result dictionary.

  @see: C{_AddSpindlesToLegacyNodeInfo} for parameter information.

  """

  if utils.storage.DiskTemplateSupportsSpaceReporting(disk_template):

    disk_info = utils.storage.LookupSpaceInfoByDiskTemplate(

        space_info, disk_template)

    result["name"] = disk_info["name"]

    result["storage_free"] = disk_info["storage_free"]

    result["storage_size"] = disk_info["storage_size"]

  else:

    # FIXME: consider displaying '-' in this case

    result["storage_free"] = 0

    result["storage_size"] = 0

def MakeLegacyNodeInfo(data, disk_template):

  """Formats the data returned by call_node_info.

  Converts the data into a single dictionary. This is fine for most use cases,

  but some require information from more than one volume group or hypervisor.

  """

  (bootid, space_info, (hv_info, )) = data

  ret = utils.JoinDisjointDicts(hv_info, {"bootid": bootid})

  _AddSpindlesToLegacyNodeInfo(ret, space_info)

  _AddStorageInfoToLegacyNodeInfoByTemplate(ret, space_info, disk_template)

  return ret

def _AnnotateDParamsDRBD(disk, (drbd_params, data_params, meta_params)):

  """Annotates just DRBD disks layouts.

  """

  assert disk.dev_type == constants.DT_DRBD8

  disk.params = objects.FillDict(drbd_params, disk.params)

  (dev_data, dev_meta) = disk.children

  dev_data.params = objects.FillDict(data_params, dev_data.params)

  dev_meta.params = objects.FillDict(meta_params, dev_meta.params)

  return disk

def _AnnotateDParamsGeneric(disk, (params, )):

  """Generic disk parameter annotation routine.

  """

  assert disk.dev_type != constants.DT_DRBD8

  disk.params = objects.FillDict(params, disk.params)

  return disk

def AnnotateDiskParams(disks, disk_params):

  """Annotates the disk objects with the disk parameters.

  @param disks: The list of disks objects to annotate

  @param disk_params: The disk parameters for annotation

  @returns: A list of disk objects annotated

  """

  def AnnotateDisk(disk):

    if disk.dev_type == constants.DT_DISKLESS:

      return disk

    ld_params = objects.Disk.ComputeLDParams(disk.dev_type, disk_params)

    if disk.dev_type == constants.DT_DRBD8:

      return _AnnotateDParamsDRBD(disk, ld_params)

    else:

      return _AnnotateDParamsGeneric(disk, ld_params)

  return [AnnotateDisk(disk.Copy()) for disk in disks]

def _GetExclusiveStorageFlag(cfg, node_uuid):

  ni = cfg.GetNodeInfo(node_uuid)

  if ni is None:

    raise errors.OpPrereqError("Invalid node name %s" % node_uuid,

                               errors.ECODE_NOENT)

  return cfg.GetNdParams(ni)[constants.ND_EXCLUSIVE_STORAGE]

def _AddExclusiveStorageFlagToLvmStorageUnits(storage_units, es_flag):

  """Adds the exclusive storage flag to lvm units.

  This function creates a copy of the storage_units lists, with the

  es_flag being added to all lvm storage units.

  @type storage_units: list of pairs (string, string)

  @param storage_units: list of 'raw' storage units, consisting only of

    (storage_type, storage_key)

  @type es_flag: boolean

  @param es_flag: exclusive storage flag

  @rtype: list of tuples (string, string, list)

  @return: list of storage units (storage_type, storage_key, params) with

    the params containing the es_flag for lvm-vg storage units

  """

  result = []

  for (storage_type, storage_key) in storage_units:

    if storage_type in [constants.ST_LVM_VG]:

      result.append((storage_type, storage_key, [es_flag]))

      if es_flag:

        result.append((constants.ST_LVM_PV, storage_key, [es_flag]))

    else:

      result.append((storage_type, storage_key, []))

  return result

def GetExclusiveStorageForNodes(cfg, node_uuids):

  """Return the exclusive storage flag for all the given nodes.

  @type cfg: L{config.ConfigWriter}

  @param cfg: cluster configuration

  @type node_uuids: list or tuple

  @param node_uuids: node UUIDs for which to read the flag

  @rtype: dict

  @return: mapping from node uuids to exclusive storage flags

  @raise errors.OpPrereqError: if any given node name has no corresponding

  node

  """

  getflag = lambda n: _GetExclusiveStorageFlag(cfg, n)

  flags = map(getflag, node_uuids)

  return dict(zip(node_uuids, flags))

def PrepareStorageUnitsForNodes(cfg, storage_units, node_uuids):

  """Return the lvm storage unit for all the given nodes.

  Main purpose of this function is to map the exclusive storage flag, which

  can be different for each node, to the default LVM storage unit.

  @type cfg: L{config.ConfigWriter}

  @param cfg: cluster configuration

  @type storage_units: list of pairs (string, string)

  @param storage_units: list of 'raw' storage units, e.g. pairs of

    (storage_type, storage_key)

  @type node_uuids: list or tuple

  @param node_uuids: node UUIDs for which to read the flag

  @rtype: dict

  @return: mapping from node uuids to a list of storage units which include

    the exclusive storage flag for lvm storage

  @raise errors.OpPrereqError: if any given node name has no corresponding