-#!/usr/bin/python
+#
#
# Copyright (C) 2006, 2007 Google Inc.
# 02110-1301, USA.
-"""Script to show add a new node to the cluster
+"""Inter-node RPC library.
"""
-# pylint: disable-msg=C0103
+# pylint: disable-msg=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 os
+import logging
+import zlib
+import base64
+
+from ganeti import utils
+from ganeti import objects
+from ganeti import http
+from ganeti import serializer
+from ganeti import constants
+from ganeti import errors
+
+# pylint has a bug here, doesn't see this import
+import ganeti.http.client # pylint: disable-msg=W0611
+
-from twisted.internet.pollreactor import PollReactor
+# Module level variable
+_http_manager = None
-class ReReactor(PollReactor):
- """A re-startable Reactor implementation.
+
+def Init():
+ """Initializes the module-global HTTP client manager.
+
+ Must be called before using any RPC function.
"""
- def run(self, installSignalHandlers=1):
- """Custom run method.
+ global _http_manager # pylint: disable-msg=W0603
- This is customized run that, before calling Reactor.run, will
- reinstall the shutdown events and re-create the threadpool in case
- these are not present (as will happen on the second run of the
- reactor).
+ assert not _http_manager, "RPC module initialized more than once"
- """
- if not 'shutdown' in self._eventTriggers:
- # the shutdown queue has been killed, we are most probably
- # at the second run, thus recreate the queue
- self.addSystemEventTrigger('during', 'shutdown', self.crash)
- self.addSystemEventTrigger('during', 'shutdown', self.disconnectAll)
- if self.threadpool is not None and self.threadpool.joined == 1:
- # in case the threadpool has been stopped, re-start it
- # and add a trigger to stop it at reactor shutdown
- self.threadpool.start()
- self.addSystemEventTrigger('during', 'shutdown', self.threadpool.stop)
+ _http_manager = http.client.HttpClientManager()
- return PollReactor.run(self, installSignalHandlers)
+def Shutdown():
+ """Stops the module-global HTTP client manager.
-import twisted.internet.main
-twisted.internet.main.installReactor(ReReactor())
+ Must be called before quitting the program.
-from twisted.spread import pb
-from twisted.internet import reactor
-from twisted.cred import credentials
-from OpenSSL import SSL, crypto
+ """
+ global _http_manager # pylint: disable-msg=W0603
+
+ if _http_manager:
+ _http_manager.Shutdown()
+ _http_manager = None
-from ganeti import logger
-from ganeti import utils
-from ganeti import errors
-from ganeti import constants
-from ganeti import objects
-from ganeti import ssconf
-class NodeController:
- """Node-handling class.
+class RpcResult(object):
+ """RPC Result class.
- For each node that we speak with, we create an instance of this
- class, so that we have a safe place to store the details of this
- individual call.
+ This class holds an RPC result. It is needed since in multi-node
+ calls we can't raise an exception just because one 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
+ @type failed: boolean
+ @ivar failed: whether the operation failed at RPC level (not
+ application level on the remote node)
+ @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
"""
- def __init__(self, parent, node):
- self.parent = parent
+ def __init__(self, data=None, failed=False, offline=False,
+ call=None, node=None):
+ self.failed = failed
+ self.offline = offline
+ self.call = call
self.node = node
+ if offline:
+ self.failed = True
+ self.error = "Node is marked offline"
+ self.data = self.payload = None
+ elif failed:
+ self.error = data
+ self.data = self.payload = None
+ else:
+ self.data = data
+ self.error = None
+ if isinstance(data, (tuple, list)) and len(data) == 2:
+ self.payload = data[1]
+ else:
+ self.payload = None
- def _check_end(self):
- """Stop the reactor if we got all the results.
+ def Raise(self):
+ """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 len(self.parent.results) == len(self.parent.nc):
- reactor.stop()
+ if self.failed:
+ raise errors.OpExecError("Call '%s' to node '%s' has failed: %s" %
+ (self.call, self.node, self.error))
+
+ def RemoteFailMsg(self):
+ """Check if the remote procedure failed.
- def cb_call(self, obj):
- """Callback for successful connect.
+ This is valid only for RPC calls which return result of the form
+ (status, data | error_msg).
- If the connect and login sequence succeeded, we proceed with
- making the actual call.
+ @return: empty string for succcess, otherwise an error message
"""
- deferred = obj.callRemote(self.parent.procedure, self.parent.args)
- deferred.addCallbacks(self.cb_done, self.cb_err2)
+ def _EnsureErr(val):
+ """Helper to ensure we return a 'True' value for error."""
+ if val:
+ return val
+ else:
+ return "No error information"
+
+ if self.failed:
+ return _EnsureErr(self.error)
+ if not isinstance(self.data, (tuple, list)):
+ return "Invalid result type (%s)" % type(self.data)
+ if len(self.data) != 2:
+ return "Invalid result length (%d), expected 2" % len(self.data)
+ if not self.data[0]:
+ return _EnsureErr(self.data[1])
+ return ""
+
+
+class Client:
+ """RPC Client class.
+
+ This class, given a (remote) method name, a list of parameters and a
+ list of nodes, will contact (in parallel) all nodes, and return a
+ dict of results (key: node name, value: result).
+
+ One current bug is that generic failure is still signaled by
+ 'False' result, which is not good. This overloading of values can
+ cause bugs.
+
+ """
+ def __init__(self, procedure, body, port):
+ self.procedure = procedure
+ self.body = body
+ self.port = port
+ self.nc = {}
+
+ self._ssl_params = \
+ http.HttpSslParams(ssl_key_path=constants.SSL_CERT_FILE,
+ ssl_cert_path=constants.SSL_CERT_FILE)
- def cb_done(self, result):
- """Callback for successful call.
+ def ConnectList(self, node_list, address_list=None):
+ """Add a list of nodes to the target nodes.
- When we receive the result from a call, we check if it was an
- error and if so we raise a generic RemoteError (we can't pass yet
- the actual exception over). If there was no error, we store the
- result.
+ @type node_list: list
+ @param node_list: the list of node names to connect
+ @type address_list: list or None
+ @keyword address_list: either None or a list with node addresses,
+ which must have the same length as the node list
"""
- tb, self.parent.results[self.node] = result
- self._check_end()
- if tb:
- raise errors.RemoteError("Remote procedure error calling %s on %s:"
- "\n%s" % (self.parent.procedure,
- self.node,
- tb))
+ if address_list is None:
+ address_list = [None for _ in node_list]
+ else:
+ assert len(node_list) == len(address_list), \
+ "Name and address lists should have the same length"
+ for node, address in zip(node_list, address_list):
+ self.ConnectNode(node, address)
- def cb_err1(self, reason):
- """Error callback for unsuccessful connect.
+ def ConnectNode(self, name, address=None):
+ """Add a node to the target list.
+
+ @type name: str
+ @param name: the node name
+ @type address: str
+ @keyword address: the node address, if known
"""
- logger.Error("caller_connect: could not connect to remote host %s,"
- " reason %s" % (self.node, reason))
- self.parent.results[self.node] = False
- self._check_end()
+ if address is None:
+ address = name
- def cb_err2(self, reason):
- """Error callback for unsuccessful call.
+ self.nc[name] = \
+ http.client.HttpClientRequest(address, self.port, http.HTTP_PUT,
+ "/%s" % self.procedure,
+ post_data=self.body,
+ ssl_params=self._ssl_params,
+ ssl_verify_peer=True)
- This is when the call didn't return anything, not even an error,
- or when it time out, etc.
+ def GetResults(self):
+ """Call nodes and return results.
+
+ @rtype: list
+ @return: List of RPC results
"""
- logger.Error("caller_call: could not call %s on node %s,"
- " reason %s" % (self.parent.procedure, self.node, reason))
- self.parent.results[self.node] = False
- self._check_end()
+ assert _http_manager, "RPC module not initialized"
+ _http_manager.ExecRequests(self.nc.values())
-class MirrorContextFactory:
- """Certificate verifier factory.
+ results = {}
- This factory creates contexts that verify if the remote end has a
- specific certificate (i.e. our own certificate).
+ for name, req in self.nc.iteritems():
+ if req.success and req.resp_status_code == http.HTTP_OK:
+ results[name] = RpcResult(data=serializer.LoadJson(req.resp_body),
+ node=name, call=self.procedure)
+ continue
- The checks we do are that the PEM dump of the certificate is the
- same as our own and (somewhat redundantly) that the SHA checksum is
- the same.
+ # TODO: Better error reporting
+ if req.error:
+ msg = req.error
+ else:
+ msg = req.resp_body
- """
- isClient = 1
+ logging.error("RPC error in %s from node %s: %s",
+ self.procedure, name, msg)
+ results[name] = RpcResult(data=msg, failed=True, node=name,
+ call=self.procedure)
+
+ return results
+
+
+class RpcRunner(object):
+ """RPC runner class"""
- def __init__(self):
- try:
- fd = open(constants.SSL_CERT_FILE, 'r')
- try:
- data = fd.read(16384)
- finally:
- fd.close()
- except EnvironmentError, err:
- raise errors.ConfigurationError("missing SSL certificate: %s" %
- str(err))
- self.mycert = crypto.load_certificate(crypto.FILETYPE_PEM, data)
- self.mypem = crypto.dump_certificate(crypto.FILETYPE_PEM, self.mycert)
- self.mydigest = self.mycert.digest('SHA')
+ def __init__(self, cfg):
+ """Initialized the rpc runner.
- def verifier(self, conn, x509, errno, err_depth, retcode):
- """Certificate verify method.
+ @type cfg: C{config.ConfigWriter}
+ @param cfg: the configuration object that will be used to get data
+ about the cluster
"""
- if self.mydigest != x509.digest('SHA'):
- return False
- if crypto.dump_certificate(crypto.FILETYPE_PEM, x509) != self.mypem:
- return False
- return True
+ self._cfg = cfg
+ self.port = utils.GetNodeDaemonPort()
- def getContext(self):
- """Context generator.
+ def _InstDict(self, instance, hvp=None, bep=None):
+ """Convert the given instance to a dict.
+
+ This is done via the instance's ToDict() method and additionally
+ we fill the hvparams with the cluster defaults.
+
+ @type instance: L{objects.Instance}
+ @param instance: an Instance object
+ @type hvp: dict or None
+ @param hvp: a dictionary with overridden hypervisor parameters
+ @type bep: dict or None
+ @param bep: a dictionary with overridden backend parameters
+ @rtype: dict
+ @return: the instance dict, with the hvparams filled with the
+ cluster defaults
"""
- context = SSL.Context(SSL.TLSv1_METHOD)
- context.set_verify(SSL.VERIFY_PEER, self.verifier)
- return context
+ idict = instance.ToDict()
+ cluster = self._cfg.GetClusterInfo()
+ idict["hvparams"] = cluster.FillHV(instance)
+ if hvp is not None:
+ idict["hvparams"].update(hvp)
+ idict["beparams"] = cluster.FillBE(instance)
+ if bep is not None:
+ idict["beparams"].update(bep)
+ return idict
+
+ def _ConnectList(self, client, node_list, call):
+ """Helper for computing node addresses.
+
+ @type client: L{ganeti.rpc.Client}
+ @param client: a C{Client} instance
+ @type node_list: list
+ @param node_list: the node list we should connect
+ @type call: string
+ @param call: the name of the remote procedure call, for filling in
+ correctly any eventual offline nodes' results
-class Client:
- """RPC Client class.
+ """
+ all_nodes = self._cfg.GetAllNodesInfo()
+ name_list = []
+ addr_list = []
+ skip_dict = {}
+ for node in node_list:
+ if node in all_nodes:
+ if all_nodes[node].offline:
+ skip_dict[node] = RpcResult(node=node, offline=True, call=call)
+ continue
+ val = all_nodes[node].primary_ip
+ else:
+ val = None
+ addr_list.append(val)
+ name_list.append(node)
+ if name_list:
+ client.ConnectList(name_list, address_list=addr_list)
+ return skip_dict
+
+ def _ConnectNode(self, client, node, call):
+ """Helper for computing one node's address.
+
+ @type client: L{ganeti.rpc.Client}
+ @param client: a C{Client} instance
+ @type node: str
+ @param node: the node we should connect
+ @type call: string
+ @param call: the name of the remote procedure call, for filling in
+ correctly any eventual offline nodes' results
- This class, given a (remote) ethod name, a list of parameters and a
- list of nodes, will contact (in parallel) all nodes, and return a
- dict of results (key: node name, value: result).
+ """
+ node_info = self._cfg.GetNodeInfo(node)
+ if node_info is not None:
+ if node_info.offline:
+ return RpcResult(node=node, offline=True, call=call)
+ addr = node_info.primary_ip
+ else:
+ addr = None
+ client.ConnectNode(node, address=addr)
- One current bug is that generic failure is still signalled by
- 'False' result, which is not good. This overloading of values can
- cause bugs.
+ def _MultiNodeCall(self, node_list, procedure, args):
+ """Helper for making a multi-node call
- """
- result_set = False
- result = False
- allresult = []
-
- def __init__(self, procedure, args):
- ss = ssconf.SimpleStore()
- self.port = ss.GetNodeDaemonPort()
- self.nodepw = ss.GetNodeDaemonPassword()
- self.nc = {}
- self.results = {}
- self.procedure = procedure
- self.args = args
+ """
+ body = serializer.DumpJson(args, indent=False)
+ c = Client(procedure, body, self.port)
+ skip_dict = self._ConnectList(c, node_list, procedure)
+ skip_dict.update(c.GetResults())
+ return skip_dict
- #--- generic connector -------------
+ @classmethod
+ def _StaticMultiNodeCall(cls, node_list, procedure, args,
+ address_list=None):
+ """Helper for making a multi-node static call
- def connect_list(self, node_list):
- """Add a list of nodes to the target nodes.
+ """
+ body = serializer.DumpJson(args, indent=False)
+ c = Client(procedure, body, utils.GetNodeDaemonPort())
+ c.ConnectList(node_list, address_list=address_list)
+ return c.GetResults()
+
+ def _SingleNodeCall(self, node, procedure, args):
+ """Helper for making a single-node call
"""
- for node in node_list:
- self.connect(node)
+ body = serializer.DumpJson(args, indent=False)
+ c = Client(procedure, body, self.port)
+ result = self._ConnectNode(c, node, procedure)
+ if result is None:
+ # we did connect, node is not offline
+ result = c.GetResults()[node]
+ return result
- def connect(self, connect_node):
- """Add a node to the target list.
+ @classmethod
+ def _StaticSingleNodeCall(cls, node, procedure, args):
+ """Helper for making a single-node static call
"""
- factory = pb.PBClientFactory()
- self.nc[connect_node] = nc = NodeController(self, connect_node)
- reactor.connectSSL(connect_node, self.port, factory,
- MirrorContextFactory())
- #d = factory.getRootObject()
- d = factory.login(credentials.UsernamePassword("master_node", self.nodepw))
- d.addCallbacks(nc.cb_call, nc.cb_err1)
+ body = serializer.DumpJson(args, indent=False)
+ c = Client(procedure, body, utils.GetNodeDaemonPort())
+ c.ConnectNode(node)
+ return c.GetResults()[node]
+
+ @staticmethod
+ def _Compress(data):
+ """Compresses a string for transport over RPC.
- def getresult(self):
- """Return the results of the call.
+ Small amounts of data are not compressed.
+
+ @type data: str
+ @param data: Data
+ @rtype: tuple
+ @return: Encoded data to send
"""
- return self.results
+ # 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)))
- def run(self):
- """Wrapper over reactor.run().
+ #
+ # Begin RPC calls
+ #
- This function simply calls reactor.run() if we have any requests
- queued, otherwise it does nothing.
+ def call_volume_list(self, node_list, vg_name):
+ """Gets the logical volumes present in a given volume group.
+
+ This is a multi-node call.
"""
- if self.nc:
- reactor.run()
+ return self._MultiNodeCall(node_list, "volume_list", [vg_name])
+ def call_vg_list(self, node_list):
+ """Gets the volume group list.
-def call_volume_list(node_list, vg_name):
- """Gets the logical volumes present in a given volume group.
+ This is a multi-node call.
- This is a multi-node call.
+ """
+ return self._MultiNodeCall(node_list, "vg_list", [])
- """
- c = Client("volume_list", [vg_name])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ def call_bridges_exist(self, node, bridges_list):
+ """Checks if a node has all the bridges given.
+ This method checks if all bridges given in the bridges_list are
+ present on the remote node, so that an instance that uses interfaces
+ on those bridges can be started.
-def call_vg_list(node_list):
- """Gets the volume group list.
+ This is a single-node call.
- This is a multi-node call.
+ """
+ return self._SingleNodeCall(node, "bridges_exist", [bridges_list])
- """
- c = Client("vg_list", [])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ def call_instance_start(self, node, instance, hvp, bep):
+ """Starts an instance.
+ This is a single-node call.
-def call_bridges_exist(node, bridges_list):
- """Checks if a node has all the bridges given.
+ """
+ idict = self._InstDict(instance, hvp=hvp, bep=bep)
+ return self._SingleNodeCall(node, "instance_start", [idict])
- This method checks if all bridges given in the bridges_list are
- present on the remote node, so that an instance that uses interfaces
- on those bridges can be started.
+ def call_instance_shutdown(self, node, instance):
+ """Stops an instance.
- This is a single-node call.
+ This is a single-node call.
- """
- c = Client("bridges_exist", [bridges_list])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "instance_shutdown",
+ [self._InstDict(instance)])
+ def call_migration_info(self, node, instance):
+ """Gather the information necessary to prepare an instance migration.
-def call_instance_start(node, instance, extra_args):
- """Stars an instance.
+ This is a single-node call.
- This is a single-node call.
+ @type node: string
+ @param node: the node on which the instance is currently running
+ @type instance: C{objects.Instance}
+ @param instance: the instance definition
- """
- c = Client("instance_start", [instance.Dumps(), extra_args])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "migration_info",
+ [self._InstDict(instance)])
+ def call_accept_instance(self, node, instance, info, target):
+ """Prepare a node to accept an instance.
-def call_instance_shutdown(node, instance):
- """Stops an instance.
+ This is a single-node call.
- This is a single-node call.
+ @type node: string
+ @param node: the target node for the migration
+ @type instance: C{objects.Instance}
+ @param instance: the instance definition
+ @type info: opaque/hypervisor specific (string/data)
+ @param info: result for the call_migration_info call
+ @type target: string
+ @param target: target hostname (usually ip address) (on the node itself)
- """
- c = Client("instance_shutdown", [instance.Dumps()])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "accept_instance",
+ [self._InstDict(instance), info, target])
+ def call_finalize_migration(self, node, instance, info, success):
+ """Finalize any target-node migration specific operation.
-def call_instance_os_add(node, inst, osdev, swapdev):
- """Installs an OS on the given instance.
+ This is called both in case of a successful migration and in case of error
+ (in which case it should abort the migration).
- This is a single-node call.
+ This is a single-node call.
- """
- params = [inst.Dumps(), osdev, swapdev]
- c = Client("instance_os_add", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ @type node: string
+ @param node: the target node for the migration
+ @type instance: C{objects.Instance}
+ @param instance: the instance definition
+ @type info: opaque/hypervisor specific (string/data)
+ @param info: result for the call_migration_info call
+ @type success: boolean
+ @param success: whether the migration was a success or a failure
+ """
+ return self._SingleNodeCall(node, "finalize_migration",
+ [self._InstDict(instance), info, success])
-def call_instance_info(node, instance):
- """Returns information about a single instance.
+ def call_instance_migrate(self, node, instance, target, live):
+ """Migrate an instance.
- This is a single-node call.
+ This is a single-node call.
- """
- c = Client("instance_info", [instance])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ @type node: string
+ @param node: the node on which the instance is currently running
+ @type instance: C{objects.Instance}
+ @param instance: the instance definition
+ @type target: string
+ @param target: the target node name
+ @type live: boolean
+ @param live: whether the migration should be done live or not (the
+ interpretation of this parameter is left to the hypervisor)
+ """
+ return self._SingleNodeCall(node, "instance_migrate",
+ [self._InstDict(instance), target, live])
-def call_all_instances_info(node_list):
- """Returns information about all instances on a given node.
+ def call_instance_reboot(self, node, instance, reboot_type):
+ """Reboots an instance.
- This is a single-node call.
+ This is a single-node call.
- """
- c = Client("all_instances_info", [])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ """
+ return self._SingleNodeCall(node, "instance_reboot",
+ [self._InstDict(instance), reboot_type])
+ def call_instance_os_add(self, node, inst):
+ """Installs an OS on the given instance.
-def call_instance_list(node_list):
- """Returns the list of running instances on a given node.
+ This is a single-node call.
- This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "instance_os_add",
+ [self._InstDict(inst)])
- """
- c = Client("instance_list", [])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ def call_instance_run_rename(self, node, inst, old_name):
+ """Run the OS rename script for an instance.
+ This is a single-node call.
-def call_node_info(node_list, vg_name):
- """Return node information.
+ """
+ return self._SingleNodeCall(node, "instance_run_rename",
+ [self._InstDict(inst), old_name])
- This will return memory information and volume group size and free
- space.
+ def call_instance_info(self, node, instance, hname):
+ """Returns information about a single instance.
- This is a multi-node call.
+ This is a single-node call.
- """
- c = Client("node_info", [vg_name])
- c.connect_list(node_list)
- c.run()
- retux = c.getresult()
+ @type node: list
+ @param node: the list of nodes to query
+ @type instance: string
+ @param instance: the instance name
+ @type hname: string
+ @param hname: the hypervisor type of the instance
- for node_name in retux:
- ret = retux.get(node_name, False)
- if type(ret) != dict:
- logger.Error("could not connect to node %s" % (node_name))
- ret = {}
+ """
+ return self._SingleNodeCall(node, "instance_info", [instance, hname])
- utils.CheckDict(ret,
- { 'memory_total' : '-',
- 'memory_dom0' : '-',
- 'memory_free' : '-',
- 'vg_size' : 'node_unreachable',
- 'vg_free' : '-' },
- "call_node_info",
- )
- return retux
+ def call_instance_migratable(self, node, instance):
+ """Checks whether the given instance can be migrated.
+ This is a single-node call.
-def call_node_add(node, dsa, dsapub, rsa, rsapub, ssh, sshpub):
- """Add a node to the cluster.
+ @param node: the node to query
+ @type instance: L{objects.Instance}
+ @param instance: the instance to check
- This is a single-node call.
- """
- params = [dsa, dsapub, rsa, rsapub, ssh, sshpub]
- c = Client("node_add", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "instance_migratable",
+ [self._InstDict(instance)])
+ def call_all_instances_info(self, node_list, hypervisor_list):
+ """Returns information about all instances on the given nodes.
-def call_node_verify(node_list, checkdict):
- """Request verification of given parameters.
+ This is a multi-node call.
- This is a multi-node call.
+ @type node_list: list
+ @param node_list: the list of nodes to query
+ @type hypervisor_list: list
+ @param hypervisor_list: the hypervisors to query for instances
- """
- c = Client("node_verify", [checkdict])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ """
+ return self._MultiNodeCall(node_list, "all_instances_info",
+ [hypervisor_list])
+ def call_instance_list(self, node_list, hypervisor_list):
+ """Returns the list of running instances on a given node.
-def call_node_start_master(node):
- """Tells a node to activate itself as a master.
+ This is a multi-node call.
- This is a single-node call.
+ @type node_list: list
+ @param node_list: the list of nodes to query
+ @type hypervisor_list: list
+ @param hypervisor_list: the hypervisors to query for instances
- """
- c = Client("node_start_master", [])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._MultiNodeCall(node_list, "instance_list", [hypervisor_list])
+ def call_node_tcp_ping(self, node, source, target, port, timeout,
+ live_port_needed):
+ """Do a TcpPing on the remote node
-def call_node_stop_master(node):
- """Tells a node to demote itself from master status.
+ This is a single-node call.
- This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "node_tcp_ping",
+ [source, target, port, timeout,
+ live_port_needed])
- """
- c = Client("node_stop_master", [])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ def call_node_has_ip_address(self, node, address):
+ """Checks if a node has the given IP address.
+ This is a single-node call.
-def call_version(node_list):
- """Query node version.
+ """
+ return self._SingleNodeCall(node, "node_has_ip_address", [address])
- This is a multi-node call.
+ def call_node_info(self, node_list, vg_name, hypervisor_type):
+ """Return node information.
- """
- c = Client("version", [])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ This will return memory information and volume group size and free
+ space.
+ This is a multi-node call.
-def call_blockdev_create(node, bdev, size, on_primary, info):
- """Request creation of a given block device.
+ @type node_list: list
+ @param node_list: the list of nodes to query
+ @type vg_name: C{string}
+ @param vg_name: the name of the volume group to ask for disk space
+ information
+ @type hypervisor_type: C{str}
+ @param hypervisor_type: the name of the hypervisor to ask for
+ memory information
- This is a single-node call.
+ """
+ retux = self._MultiNodeCall(node_list, "node_info",
+ [vg_name, hypervisor_type])
+
+ for result in retux.itervalues():
+ if result.failed or not isinstance(result.data, dict):
+ result.data = {}
+ if result.offline:
+ log_name = None
+ else:
+ log_name = "call_node_info"
+
+ utils.CheckDict(result.data, {
+ 'memory_total' : '-',
+ 'memory_dom0' : '-',
+ 'memory_free' : '-',
+ 'vg_size' : 'node_unreachable',
+ 'vg_free' : '-',
+ }, log_name)
+ return retux
+
+ def call_node_add(self, node, dsa, dsapub, rsa, rsapub, ssh, sshpub):
+ """Add a node to the cluster.
+
+ This is a single-node call.
- """
- params = [bdev.Dumps(), size, on_primary, info]
- c = Client("blockdev_create", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "node_add",
+ [dsa, dsapub, rsa, rsapub, ssh, sshpub])
+ def call_node_verify(self, node_list, checkdict, cluster_name):
+ """Request verification of given parameters.
-def call_blockdev_remove(node, bdev):
- """Request removal of a given block device.
+ This is a multi-node call.
- This is a single-node call.
+ """
+ return self._MultiNodeCall(node_list, "node_verify",
+ [checkdict, cluster_name])
- """
- c = Client("blockdev_remove", [bdev.Dumps()])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ @classmethod
+ def call_node_start_master(cls, node, start_daemons, no_voting):
+ """Tells a node to activate itself as a master.
+ This is a single-node call.
-def call_blockdev_assemble(node, disk, on_primary):
- """Request assembling of a given block device.
+ """
+ return cls._StaticSingleNodeCall(node, "node_start_master",
+ [start_daemons, no_voting])
- This is a single-node call.
+ @classmethod
+ def call_node_stop_master(cls, node, stop_daemons):
+ """Tells a node to demote itself from master status.
- """
- params = [disk.Dumps(), on_primary]
- c = Client("blockdev_assemble", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ This is a single-node call.
+ """
+ return cls._StaticSingleNodeCall(node, "node_stop_master", [stop_daemons])
-def call_blockdev_shutdown(node, disk):
- """Request shutdown of a given block device.
+ @classmethod
+ def call_master_info(cls, node_list):
+ """Query master info.
- This is a single-node call.
+ This is a multi-node call.
- """
- c = Client("blockdev_shutdown", [disk.Dumps()])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ # TODO: should this method query down nodes?
+ return cls._StaticMultiNodeCall(node_list, "master_info", [])
+ def call_version(self, node_list):
+ """Query node version.
-def call_blockdev_addchild(node, bdev, ndev):
- """Request adding a new child to a (mirroring) device.
+ This is a multi-node call.
- This is a single-node call.
+ """
+ return self._MultiNodeCall(node_list, "version", [])
- """
- params = [bdev.Dumps(), ndev.Dumps()]
- c = Client("blockdev_addchild", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ def call_blockdev_create(self, node, bdev, size, owner, on_primary, info):
+ """Request creation of a given block device.
+ This is a single-node call.
-def call_blockdev_removechild(node, bdev, ndev):
- """Request removing a new child from a (mirroring) device.
+ """
+ return self._SingleNodeCall(node, "blockdev_create",
+ [bdev.ToDict(), size, owner, on_primary, info])
- This is a single-node call.
+ def call_blockdev_remove(self, node, bdev):
+ """Request removal of a given block device.
- """
- params = [bdev.Dumps(), ndev.Dumps()]
- c = Client("blockdev_removechild", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "blockdev_remove", [bdev.ToDict()])
-def call_blockdev_getmirrorstatus(node, disks):
- """Request status of a (mirroring) device.
+ def call_blockdev_rename(self, node, devlist):
+ """Request rename of the given block devices.
- This is a single-node call.
+ This is a single-node call.
- """
- params = [dsk.Dumps() for dsk in disks]
- c = Client("blockdev_getmirrorstatus", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "blockdev_rename",
+ [(d.ToDict(), uid) for d, uid in devlist])
+ def call_blockdev_assemble(self, node, disk, owner, on_primary):
+ """Request assembling of a given block device.
-def call_blockdev_find(node, disk):
- """Request identification of a given block device.
+ This is a single-node call.
- This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "blockdev_assemble",
+ [disk.ToDict(), owner, on_primary])
- """
- c = Client("blockdev_find", [disk.Dumps()])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ def call_blockdev_shutdown(self, node, disk):
+ """Request shutdown of a given block device.
+ This is a single-node call.
-def call_upload_file(node_list, file_name):
- """Upload a file.
+ """
+ return self._SingleNodeCall(node, "blockdev_shutdown", [disk.ToDict()])
- The node will refuse the operation in case the file is not on the
- approved file list.
+ def call_blockdev_addchildren(self, node, bdev, ndevs):
+ """Request adding a list of children to a (mirroring) device.
- This is a multi-node call.
+ This is a single-node call.
- """
- fh = file(file_name)
- try:
- data = fh.read()
- finally:
- fh.close()
- st = os.stat(file_name)
- params = [file_name, data, st.st_mode, st.st_uid, st.st_gid,
- st.st_atime, st.st_mtime]
- c = Client("upload_file", params)
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ """
+ return self._SingleNodeCall(node, "blockdev_addchildren",
+ [bdev.ToDict(),
+ [disk.ToDict() for disk in ndevs]])
+ def call_blockdev_removechildren(self, node, bdev, ndevs):
+ """Request removing a list of children from a (mirroring) device.
-def call_os_diagnose(node_list):
- """Request a diagnose of OS definitions.
+ This is a single-node call.
- This is a multi-node call.
+ """
+ return self._SingleNodeCall(node, "blockdev_removechildren",
+ [bdev.ToDict(),
+ [disk.ToDict() for disk in ndevs]])
- """
- c = Client("os_diagnose", [])
- c.connect_list(node_list)
- c.run()
- result = c.getresult()
- new_result = {}
- for node_name in result:
- nr = []
- if result[node_name]:
- for data in result[node_name]:
- if data:
- if isinstance(data, basestring):
- nr.append(objects.ConfigObject.Loads(data))
- elif isinstance(data, tuple) and len(data) == 2:
- nr.append(errors.InvalidOS(data[0], data[1]))
- else:
- raise errors.ProgrammerError("Invalid data from"
- " xcserver.os_diagnose")
- new_result[node_name] = nr
- return new_result
-
-
-def call_os_get(node_list, name):
- """Returns an OS definition.
-
- This is a multi-node call.
+ def call_blockdev_getmirrorstatus(self, node, disks):
+ """Request status of a (mirroring) device.
- """
- c = Client("os_get", [name])
- c.connect_list(node_list)
- c.run()
- result = c.getresult()
- new_result = {}
- for node_name in result:
- data = result[node_name]
- if isinstance(data, basestring):
- new_result[node_name] = objects.ConfigObject.Loads(data)
- elif isinstance(data, tuple) and len(data) == 2:
- new_result[node_name] = errors.InvalidOS(data[0], data[1])
- else:
- new_result[node_name] = data
- return new_result
+ This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "blockdev_getmirrorstatus",
+ [dsk.ToDict() for dsk in disks])
-def call_hooks_runner(node_list, hpath, phase, env):
- """Call the hooks runner.
+ def call_blockdev_find(self, node, disk):
+ """Request identification of a given block device.
- Args:
- - op: the OpCode instance
- - env: a dictionary with the environment
+ This is a single-node call.
- This is a multi-node call.
+ """
+ return self._SingleNodeCall(node, "blockdev_find", [disk.ToDict()])
- """
- params = [hpath, phase, env]
- c = Client("hooks_runner", params)
- c.connect_list(node_list)
- c.run()
- result = c.getresult()
- return result
+ def call_blockdev_close(self, node, instance_name, disks):
+ """Closes the given block devices.
+ This is a single-node call.
-def call_blockdev_snapshot(node, cf_bdev):
- """Request a snapshot of the given block device.
+ """
+ params = [instance_name, [cf.ToDict() for cf in disks]]
+ return self._SingleNodeCall(node, "blockdev_close", params)
- This is a single-node call.
+ def call_blockdev_getsizes(self, node, disks):
+ """Returns the size of the given disks.
- """
- c = Client("blockdev_snapshot", [cf_bdev.Dumps()])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ This is a single-node call.
+ """
+ params = [[cf.ToDict() for cf in disks]]
+ return self._SingleNodeCall(node, "blockdev_getsize", params)
-def call_snapshot_export(node, snap_bdev, dest_node, instance):
- """Request the export of a given snapshot.
+ def call_drbd_disconnect_net(self, node_list, nodes_ip, disks):
+ """Disconnects the network of the given drbd devices.
- This is a single-node call.
+ This is a multi-node call.
- """
- params = [snap_bdev.Dumps(), dest_node, instance.Dumps()]
- c = Client("snapshot_export", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._MultiNodeCall(node_list, "drbd_disconnect_net",
+ [nodes_ip, [cf.ToDict() for cf in disks]])
+
+ def call_drbd_attach_net(self, node_list, nodes_ip,
+ disks, instance_name, multimaster):
+ """Disconnects the given drbd devices.
+ This is a multi-node call.
-def call_finalize_export(node, instance, snap_disks):
- """Request the completion of an export operation.
+ """
+ return self._MultiNodeCall(node_list, "drbd_attach_net",
+ [nodes_ip, [cf.ToDict() for cf in disks],
+ instance_name, multimaster])
- This writes the export config file, etc.
+ def call_drbd_wait_sync(self, node_list, nodes_ip, disks):
+ """Waits for the synchronization of drbd devices is complete.
- This is a single-node call.
+ This is a multi-node call.
- """
- flat_disks = []
- for disk in snap_disks:
- flat_disks.append(disk.Dumps())
- params = [instance.Dumps(), flat_disks]
- c = Client("finalize_export", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._MultiNodeCall(node_list, "drbd_wait_sync",
+ [nodes_ip, [cf.ToDict() for cf in disks]])
+ @classmethod
+ def call_upload_file(cls, node_list, file_name, address_list=None):
+ """Upload a file.
-def call_export_info(node, path):
- """Queries the export information in a given path.
+ The node will refuse the operation in case the file is not on the
+ approved file list.
- This is a single-node call.
+ This is a multi-node call.
- """
- c = Client("export_info", [path])
- c.connect(node)
- c.run()
- result = c.getresult().get(node, False)
- if not result:
+ @type node_list: list
+ @param node_list: the list of node names to upload to
+ @type file_name: str
+ @param file_name: the filename to upload
+ @type address_list: list or None
+ @keyword address_list: an optional list of node addresses, in order
+ to optimize the RPC speed
+
+ """
+ file_contents = utils.ReadFile(file_name)
+ data = cls._Compress(file_contents)
+ st = os.stat(file_name)
+ params = [file_name, data, st.st_mode, st.st_uid, st.st_gid,
+ st.st_atime, st.st_mtime]
+ return cls._StaticMultiNodeCall(node_list, "upload_file", params,
+ address_list=address_list)
+
+ @classmethod
+ def call_write_ssconf_files(cls, node_list, values):
+ """Write ssconf files.
+
+ This is a multi-node call.
+
+ """
+ return cls._StaticMultiNodeCall(node_list, "write_ssconf_files", [values])
+
+ def call_os_diagnose(self, node_list):
+ """Request a diagnose of OS definitions.
+
+ This is a multi-node call.
+
+ """
+ result = self._MultiNodeCall(node_list, "os_diagnose", [])
+
+ for node_result in result.values():
+ if not node_result.failed and node_result.data:
+ node_result.data = [objects.OS.FromDict(oss)
+ for oss in node_result.data]
return result
- return objects.SerializableConfigParser.Loads(result)
+ def call_os_get(self, node, name):
+ """Returns an OS definition.
-def call_instance_os_import(node, inst, osdev, swapdev, src_node, src_image):
- """Request the import of a backup into an instance.
+ This is a single-node call.
- This is a single-node call.
+ """
+ result = self._SingleNodeCall(node, "os_get", [name])
+ if not result.failed and isinstance(result.data, dict):
+ result.data = objects.OS.FromDict(result.data)
+ return result
- """
- params = [inst.Dumps(), osdev, swapdev, src_node, src_image]
- c = Client("instance_os_import", params)
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ def call_hooks_runner(self, node_list, hpath, phase, env):
+ """Call the hooks runner.
+ Args:
+ - op: the OpCode instance
+ - env: a dictionary with the environment
-def call_export_list(node_list):
- """Gets the stored exports list.
+ This is a multi-node call.
- This is a multi-node call.
+ """
+ params = [hpath, phase, env]
+ return self._MultiNodeCall(node_list, "hooks_runner", params)
- """
- c = Client("export_list", [])
- c.connect_list(node_list)
- c.run()
- result = c.getresult()
- return result
+ def call_iallocator_runner(self, node, name, idata):
+ """Call an iallocator on a remote node
+ Args:
+ - name: the iallocator name
+ - input: the json-encoded input string
-def call_export_remove(node, export):
- """Requests removal of a given export.
+ This is a single-node call.
- This is a single-node call.
+ """
+ return self._SingleNodeCall(node, "iallocator_runner", [name, idata])
- """
- c = Client("export_remove", [export])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ def call_blockdev_grow(self, node, cf_bdev, amount):
+ """Request a snapshot of the given block device.
+ This is a single-node call.
-def call_node_leave_cluster(node):
- """Requests a node to clean the cluster information it has.
+ """
+ return self._SingleNodeCall(node, "blockdev_grow",
+ [cf_bdev.ToDict(), amount])
- This will remove the configuration information from the ganeti data
- dir.
+ def call_blockdev_snapshot(self, node, cf_bdev):
+ """Request a snapshot of the given block device.
- This is a single-node call.
+ This is a single-node call.
- """
- c = Client("node_leave_cluster", [])
- c.connect(node)
- c.run()
- return c.getresult().get(node, False)
+ """
+ return self._SingleNodeCall(node, "blockdev_snapshot", [cf_bdev.ToDict()])
+
+ def call_snapshot_export(self, node, snap_bdev, dest_node, instance,
+ cluster_name, idx):
+ """Request the export of a given snapshot.
+ This is a single-node call.
-def call_node_volumes(node_list):
- """Gets all volumes on node(s).
+ """
+ return self._SingleNodeCall(node, "snapshot_export",
+ [snap_bdev.ToDict(), dest_node,
+ self._InstDict(instance), cluster_name, idx])
- This is a multi-node call.
+ def call_finalize_export(self, node, instance, snap_disks):
+ """Request the completion of an export operation.
- """
- c = Client("node_volumes", [])
- c.connect_list(node_list)
- c.run()
- return c.getresult()
+ This writes the export config file, etc.
+
+ This is a single-node call.
+
+ """
+ flat_disks = []
+ for disk in snap_disks:
+ if isinstance(disk, bool):
+ flat_disks.append(disk)
+ else:
+ flat_disks.append(disk.ToDict())
+
+ return self._SingleNodeCall(node, "finalize_export",
+ [self._InstDict(instance), flat_disks])
+
+ def call_export_info(self, node, path):
+ """Queries the export information in a given path.
+
+ This is a single-node call.
+
+ """
+ result = self._SingleNodeCall(node, "export_info", [path])
+ if not result.failed and result.data:
+ result.data = objects.SerializableConfigParser.Loads(str(result.data))
+ return result
+
+ def call_instance_os_import(self, node, inst, src_node, src_images,
+ cluster_name):
+ """Request the import of a backup into an instance.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "instance_os_import",
+ [self._InstDict(inst), src_node, src_images,
+ cluster_name])
+
+ def call_export_list(self, node_list):
+ """Gets the stored exports list.
+
+ This is a multi-node call.
+
+ """
+ return self._MultiNodeCall(node_list, "export_list", [])
+
+ def call_export_remove(self, node, export):
+ """Requests removal of a given export.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "export_remove", [export])
+
+ @classmethod
+ def call_node_leave_cluster(cls, node):
+ """Requests a node to clean the cluster information it has.
+
+ This will remove the configuration information from the ganeti data
+ dir.
+
+ This is a single-node call.
+
+ """
+ return cls._StaticSingleNodeCall(node, "node_leave_cluster", [])
+
+ def call_node_volumes(self, node_list):
+ """Gets all volumes on node(s).
+
+ This is a multi-node call.
+
+ """
+ return self._MultiNodeCall(node_list, "node_volumes", [])
+
+ def call_node_demote_from_mc(self, node):
+ """Demote a node from the master candidate role.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "node_demote_from_mc", [])
+
+ def call_test_delay(self, node_list, duration):
+ """Sleep for a fixed time on given node(s).
+
+ This is a multi-node call.
+
+ """
+ return self._MultiNodeCall(node_list, "test_delay", [duration])
+
+ def call_file_storage_dir_create(self, node, file_storage_dir):
+ """Create the given file storage directory.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "file_storage_dir_create",
+ [file_storage_dir])
+
+ def call_file_storage_dir_remove(self, node, file_storage_dir):
+ """Remove the given file storage directory.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "file_storage_dir_remove",
+ [file_storage_dir])
+
+ def call_file_storage_dir_rename(self, node, old_file_storage_dir,
+ new_file_storage_dir):
+ """Rename file storage directory.
+
+ This is a single-node call.
+
+ """
+ return self._SingleNodeCall(node, "file_storage_dir_rename",
+ [old_file_storage_dir, new_file_storage_dir])
+
+ @classmethod
+ def call_jobqueue_update(cls, node_list, address_list, file_name, content):
+ """Update job queue.
+
+ This is a multi-node call.
+
+ """
+ return cls._StaticMultiNodeCall(node_list, "jobqueue_update",
+ [file_name, cls._Compress(content)],
+ address_list=address_list)
+
+ @classmethod
+ def call_jobqueue_purge(cls, node):
+ """Purge job queue.
+
+ This is a single-node call.
+
+ """
+ return cls._StaticSingleNodeCall(node, "jobqueue_purge", [])
+
+ @classmethod
+ def call_jobqueue_rename(cls, node_list, address_list, rename):
+ """Rename a job queue file.
+
+ This is a multi-node call.
+
+ """
+ return cls._StaticMultiNodeCall(node_list, "jobqueue_rename", rename,
+ address_list=address_list)
+
+ @classmethod
+ def call_jobqueue_set_drain(cls, node_list, drain_flag):
+ """Set the drain flag on the queue.
+
+ This is a multi-node call.
+
+ @type node_list: list
+ @param node_list: the list of nodes to query
+ @type drain_flag: bool
+ @param drain_flag: if True, will set the drain flag, otherwise reset it.
+
+ """
+ return cls._StaticMultiNodeCall(node_list, "jobqueue_set_drain",
+ [drain_flag])
+
+ def call_hypervisor_validate_params(self, node_list, hvname, hvparams):
+ """Validate the hypervisor params.
+
+ This is a multi-node call.
+
+ @type node_list: list
+ @param node_list: the list of nodes to query
+ @type hvname: string
+ @param hvname: the hypervisor name
+ @type hvparams: dict
+ @param hvparams: the hypervisor parameters to be validated
+
+ """
+ cluster = self._cfg.GetClusterInfo()
+ hv_full = cluster.FillDict(cluster.hvparams.get(hvname, {}), hvparams)
+ return self._MultiNodeCall(node_list, "hypervisor_validate_params",
+ [hvname, hv_full])