Fix 'hvparams' of '_InstanceStartupMemory' on hypervisors
[ganeti-local] / lib / backend.py
index b950910..e4a26ab 100644 (file)
@@ -56,6 +56,7 @@ from ganeti import hypervisor
 from ganeti import constants
 from ganeti.storage import bdev
 from ganeti.storage import drbd
+from ganeti.storage import filestorage
 from ganeti import objects
 from ganeti import ssconf
 from ganeti import serializer
@@ -85,7 +86,7 @@ _IES_PID_FILE = "pid"
 _IES_CA_FILE = "ca"
 
 #: Valid LVS output line regex
-_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
+_LVSLINE_REGEX = re.compile(r"^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
 
 # Actions for the master setup script
 _MASTER_START = "start"
@@ -572,12 +573,62 @@ def LeaveCluster(modify_ssh_setup):
   raise errors.QuitGanetiException(True, "Shutdown scheduled")
 
 
-def _GetVgInfo(name, excl_stor):
+def _CheckStorageParams(params, num_params):
+  """Performs sanity checks for storage parameters.
+
+  @type params: list
+  @param params: list of storage parameters
+  @type num_params: int
+  @param num_params: expected number of parameters
+
+  """
+  if params is None:
+    raise errors.ProgrammerError("No storage parameters for storage"
+                                 " reporting is provided.")
+  if not isinstance(params, list):
+    raise errors.ProgrammerError("The storage parameters are not of type"
+                                 " list: '%s'" % params)
+  if not len(params) == num_params:
+    raise errors.ProgrammerError("Did not receive the expected number of"
+                                 "storage parameters: expected %s,"
+                                 " received '%s'" % (num_params, len(params)))
+
+
+def _CheckLvmStorageParams(params):
+  """Performs sanity check for the 'exclusive storage' flag.
+
+  @see: C{_CheckStorageParams}
+
+  """
+  _CheckStorageParams(params, 1)
+  excl_stor = params[0]
+  if not isinstance(params[0], bool):
+    raise errors.ProgrammerError("Exclusive storage parameter is not"
+                                 " boolean: '%s'." % excl_stor)
+  return excl_stor
+
+
+def _GetLvmVgSpaceInfo(name, params):
+  """Wrapper around C{_GetVgInfo} which checks the storage parameters.
+
+  @type name: string
+  @param name: name of the volume group
+  @type params: list
+  @param params: list of storage parameters, which in this case should be
+    containing only one for exclusive storage
+
+  """
+  excl_stor = _CheckLvmStorageParams(params)
+  return _GetVgInfo(name, excl_stor)
+
+
+def _GetVgInfo(
+    name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
   """Retrieves information about a LVM volume group.
 
   """
   # TODO: GetVGInfo supports returning information for multiple VGs at once
-  vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
+  vginfo = info_fn([name], excl_stor)
   if vginfo:
     vg_free = int(round(vginfo[0][0], 0))
     vg_size = int(round(vginfo[0][1], 0))
@@ -586,13 +637,25 @@ def _GetVgInfo(name, excl_stor):
     vg_size = None
 
   return {
+    "type": constants.ST_LVM_VG,
     "name": name,
-    "vg_free": vg_free,
-    "vg_size": vg_size,
+    "storage_free": vg_free,
+    "storage_size": vg_size,
     }
 
 
-def _GetVgSpindlesInfo(name, excl_stor):
+def _GetLvmPvSpaceInfo(name, params):
+  """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
+
+  @see: C{_GetLvmVgSpaceInfo}
+
+  """
+  excl_stor = _CheckLvmStorageParams(params)
+  return _GetVgSpindlesInfo(name, excl_stor)
+
+
+def _GetVgSpindlesInfo(
+    name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
   """Retrieves information about spindles in an LVM volume group.
 
   @type name: string
@@ -605,14 +668,15 @@ def _GetVgSpindlesInfo(name, excl_stor):
 
   """
   if excl_stor:
-    (vg_free, vg_size) = bdev.LogicalVolume.GetVgSpindlesInfo(name)
+    (vg_free, vg_size) = info_fn(name)
   else:
     vg_free = 0
     vg_size = 0
   return {
+    "type": constants.ST_LVM_PV,
     "name": name,
-    "vg_free": vg_free,
-    "vg_size": vg_size,
+    "storage_free": vg_free,
+    "storage_size": vg_size,
     }
 
 
@@ -665,17 +729,16 @@ def _GetNamedNodeInfo(names, fn):
     return map(fn, names)
 
 
-def GetNodeInfo(storage_units, hv_specs, excl_stor):
+def GetNodeInfo(storage_units, hv_specs):
   """Gives back a hash with different information about the node.
 
-  @type storage_units: list of pairs (string, string)
-  @param storage_units: List of pairs (storage unit, identifier) to ask for disk
-                        space information. In case of lvm-vg, the identifier is
-                        the VG name.
+  @type storage_units: list of tuples (string, string, list)
+  @param storage_units: List of tuples (storage unit, identifier, parameters) to
+    ask for disk space information. In case of lvm-vg, the identifier is
+    the VG name. The parameters can contain additional, storage-type-specific
+    parameters, for example exclusive storage for lvm storage.
   @type hv_specs: list of pairs (string, dict of strings)
   @param hv_specs: list of pairs of a hypervisor's name and its hvparams
-  @type excl_stor: boolean
-  @param excl_stor: Whether exclusive_storage is active
   @rtype: tuple; (string, None/dict, None/dict)
   @return: Tuple containing boot ID, volume group information and hypervisor
     information
@@ -684,21 +747,35 @@ def GetNodeInfo(storage_units, hv_specs, excl_stor):
   bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
   storage_info = _GetNamedNodeInfo(
     storage_units,
-    (lambda storage_unit: _ApplyStorageInfoFunction(storage_unit[0],
-                                                    storage_unit[1],
-                                                    excl_stor)))
+    (lambda (storage_type, storage_key, storage_params):
+        _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
   hv_info = _GetHvInfoAll(hv_specs)
   return (bootid, storage_info, hv_info)
 
 
+def _GetFileStorageSpaceInfo(path, params):
+  """Wrapper around filestorage.GetSpaceInfo.
+
+  The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
+  and ignore the *args parameter to not leak it into the filestorage
+  module's code.
+
+  @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
+    parameters.
+
+  """
+  _CheckStorageParams(params, 0)
+  return filestorage.GetFileStorageSpaceInfo(path)
+
+
 # FIXME: implement storage reporting for all missing storage types.
 _STORAGE_TYPE_INFO_FN = {
   constants.ST_BLOCK: None,
   constants.ST_DISKLESS: None,
   constants.ST_EXT: None,
-  constants.ST_FILE: None,
-  constants.ST_LVM_PV: _GetVgSpindlesInfo,
-  constants.ST_LVM_VG: _GetVgInfo,
+  constants.ST_FILE: _GetFileStorageSpaceInfo,
+  constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
+  constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
   constants.ST_RADOS: None,
 }
 
@@ -1053,9 +1130,21 @@ def VerifyNode(what, cluster_name, all_hvparams):
                                     for bridge in what[constants.NV_BRIDGES]
                                     if not utils.BridgeExists(bridge)]
 
-  if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
-    result[constants.NV_FILE_STORAGE_PATHS] = \
-      bdev.ComputeWrongFileStoragePaths()
+  if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
+    result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
+        filestorage.ComputeWrongFileStoragePaths()
+
+  if what.get(constants.NV_FILE_STORAGE_PATH):
+    pathresult = filestorage.CheckFileStoragePath(
+        what[constants.NV_FILE_STORAGE_PATH])
+    if pathresult:
+      result[constants.NV_FILE_STORAGE_PATH] = pathresult
+
+  if what.get(constants.NV_SHARED_FILE_STORAGE_PATH):
+    pathresult = filestorage.CheckFileStoragePath(
+        what[constants.NV_SHARED_FILE_STORAGE_PATH])
+    if pathresult:
+      result[constants.NV_SHARED_FILE_STORAGE_PATH] = pathresult
 
   return result
 
@@ -1273,13 +1362,15 @@ def GetInstanceList(hypervisor_list, all_hvparams=None,
   return results
 
 
-def GetInstanceInfo(instance, hname):
+def GetInstanceInfo(instance, hname, hvparams=None):
   """Gives back the information about an instance as a dictionary.
 
   @type instance: string
   @param instance: the instance name
   @type hname: string
   @param hname: the hypervisor type of the instance
+  @type hvparams: dict of strings
+  @param hvparams: the instance's hvparams
 
   @rtype: dict
   @return: dictionary with the following keys:
@@ -1291,7 +1382,8 @@ def GetInstanceInfo(instance, hname):
   """
   output = {}
 
-  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
+  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
+                                                          hvparams=hvparams)
   if iinfo is not None:
     output["memory"] = iinfo[2]
     output["vcpus"] = iinfo[3]
@@ -1325,7 +1417,7 @@ def GetInstanceMigratable(instance):
                       iname, link_name, idx)
 
 
-def GetAllInstancesInfo(hypervisor_list):
+def GetAllInstancesInfo(hypervisor_list, all_hvparams):
   """Gather data about all instances.
 
   This is the equivalent of L{GetInstanceInfo}, except that it
@@ -1334,6 +1426,8 @@ def GetAllInstancesInfo(hypervisor_list):
 
   @type hypervisor_list: list
   @param hypervisor_list: list of hypervisors to query for instance data
+  @type all_hvparams: dict of dict of strings
+  @param all_hvparams: mapping of hypervisor names to hvparams
 
   @rtype: dict
   @return: dictionary of instance: data, with data having the following keys:
@@ -1346,7 +1440,8 @@ def GetAllInstancesInfo(hypervisor_list):
   output = {}
 
   for hname in hypervisor_list:
-    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
+    hvparams = all_hvparams[hname]
+    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
     if iinfo:
       for name, _, memory, vcpus, state, times in iinfo:
         value = {
@@ -1783,9 +1878,11 @@ def FinalizeMigrationDst(instance, info, success):
     _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
 
 
-def MigrateInstance(instance, target, live):
+def MigrateInstance(cluster_name, instance, target, live):
   """Migrates an instance to another node.
 
+  @type cluster_name: string
+  @param cluster_name: name of the cluster
   @type instance: L{objects.Instance}
   @param instance: the instance definition
   @type target: string
@@ -1799,7 +1896,7 @@ def MigrateInstance(instance, target, live):
   hyper = hypervisor.GetHypervisor(instance.hypervisor)
 
   try:
-    hyper.MigrateInstance(instance, target, live)
+    hyper.MigrateInstance(cluster_name, instance, target, live)
   except errors.HypervisorError, err:
     _Fail("Failed to migrate instance: %s", err, exc=True)
 
@@ -2340,13 +2437,13 @@ def BlockdevGetdimensions(disks):
   return result
 
 
-def BlockdevExport(disk, dest_node, dest_path, cluster_name):
+def BlockdevExport(disk, dest_node_ip, dest_path, cluster_name):
   """Export a block device to a remote node.
 
   @type disk: L{objects.Disk}
   @param disk: the description of the disk to export
-  @type dest_node: str
-  @param dest_node: the destination node to export to
+  @type dest_node_ip: str
+  @param dest_node_ip: the destination node IP to export to
   @type dest_path: str
   @param dest_path: the destination path on the target node
   @type cluster_name: str
@@ -2370,7 +2467,7 @@ def BlockdevExport(disk, dest_node, dest_path, cluster_name):
   destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
                                 " oflag=dsync", dest_path)
 
-  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
+  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node_ip,
                                                    constants.SSH_LOGIN_USER,
                                                    destcmd)
 
@@ -2742,18 +2839,24 @@ def OSEnvironment(instance, inst_os, debug=0):
     real_disk = _OpenRealBD(disk)
     result["DISK_%d_PATH" % idx] = real_disk.dev_path
     result["DISK_%d_ACCESS" % idx] = disk.mode
+    result["DISK_%d_UUID" % idx] = disk.uuid
+    if disk.name:
+      result["DISK_%d_NAME" % idx] = disk.name
     if constants.HV_DISK_TYPE in instance.hvparams:
       result["DISK_%d_FRONTEND_TYPE" % idx] = \
         instance.hvparams[constants.HV_DISK_TYPE]
-    if disk.dev_type in constants.LDS_BLOCK:
+    if disk.dev_type in constants.DTS_BLOCK:
       result["DISK_%d_BACKEND_TYPE" % idx] = "block"
-    elif disk.dev_type == constants.LD_FILE:
+    elif disk.dev_type in [constants.DT_FILE, constants.DT_SHARED_FILE]:
       result["DISK_%d_BACKEND_TYPE" % idx] = \
         "file:%s" % disk.physical_id[0]
 
   # NICs
   for idx, nic in enumerate(instance.nics):
     result["NIC_%d_MAC" % idx] = nic.mac
+    result["NIC_%d_UUID" % idx] = nic.uuid
+    if nic.name:
+      result["NIC_%d_NAME" % idx] = nic.name
     if nic.ip:
       result["NIC_%d_IP" % idx] = nic.ip
     result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
@@ -2821,7 +2924,7 @@ def DiagnoseExtStorage(top_dirs=None):
   return result
 
 
-def BlockdevGrow(disk, amount, dryrun, backingstore):
+def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
   """Grow a stack of block devices.
 
   This function is called recursively, with the childrens being the
@@ -2838,6 +2941,8 @@ def BlockdevGrow(disk, amount, dryrun, backingstore):
       only, or on "logical" storage only; e.g. DRBD is logical storage,
       whereas LVM, file, RBD are backing storage
   @rtype: (status, result)
+  @type excl_stor: boolean
+  @param excl_stor: Whether exclusive_storage is active
   @return: a tuple with the status of the operation (True/False), and
       the errors message if status is False
 
@@ -2847,7 +2952,7 @@ def BlockdevGrow(disk, amount, dryrun, backingstore):
     _Fail("Cannot find block device %s", disk)
 
   try:
-    r_dev.Grow(amount, dryrun, backingstore)
+    r_dev.Grow(amount, dryrun, backingstore, excl_stor)
   except errors.BlockDeviceError, err:
     _Fail("Failed to grow block device: %s", err, exc=True)
 
@@ -2864,12 +2969,12 @@ def BlockdevSnapshot(disk):
   @return: snapshot disk ID as (vg, lv)
 
   """
-  if disk.dev_type == constants.LD_DRBD8:
+  if disk.dev_type == constants.DT_DRBD8:
     if not disk.children:
       _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
             disk.unique_id)
     return BlockdevSnapshot(disk.children[0])
-  elif disk.dev_type == constants.LD_LV:
+  elif disk.dev_type == constants.DT_PLAIN:
     r_dev = _RecursiveFindBD(disk)
     if r_dev is not None:
       # FIXME: choose a saner value for the snapshot size
@@ -3102,11 +3207,7 @@ def _TransformFileStorageDir(fs_dir):
   @return: the normalized path if valid, None otherwise
 
   """
-  if not (constants.ENABLE_FILE_STORAGE or
-          constants.ENABLE_SHARED_FILE_STORAGE):
-    _Fail("File storage disabled at configure time")
-
-  bdev.CheckFileStoragePath(fs_dir)
+  filestorage.CheckFileStoragePath(fs_dir)
 
   return os.path.normpath(fs_dir)
 
@@ -3769,14 +3870,19 @@ def CleanupImportExport(name):
   shutil.rmtree(status_dir, ignore_errors=True)
 
 
-def _FindDisks(nodes_ip, disks):
-  """Sets the physical ID on disks and returns the block devices.
+def _SetPhysicalId(target_node_uuid, nodes_ip, disks):
+  """Sets the correct physical ID on all passed disks.
 
   """
-  # set the correct physical ID
-  my_name = netutils.Hostname.GetSysName()
   for cf in disks:
-    cf.SetPhysicalID(my_name, nodes_ip)
+    cf.SetPhysicalID(target_node_uuid, nodes_ip)
+
+
+def _FindDisks(target_node_uuid, nodes_ip, disks):
+  """Sets the physical ID on disks and returns the block devices.
+
+  """
+  _SetPhysicalId(target_node_uuid, nodes_ip, disks)
 
   bdevs = []
 
@@ -3788,11 +3894,11 @@ def _FindDisks(nodes_ip, disks):
   return bdevs
 
 
-def DrbdDisconnectNet(nodes_ip, disks):
+def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
   """Disconnects the network on a list of drbd devices.
 
   """
-  bdevs = _FindDisks(nodes_ip, disks)
+  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
 
   # disconnect disks
   for rd in bdevs:
@@ -3803,11 +3909,12 @@ def DrbdDisconnectNet(nodes_ip, disks):
             err, exc=True)
 
 
-def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
+def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
+                  multimaster):
   """Attaches the network on a list of drbd devices.
 
   """
-  bdevs = _FindDisks(nodes_ip, disks)
+  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
 
   if multimaster:
     for idx, rd in enumerate(bdevs):
@@ -3835,8 +3942,20 @@ def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
     for rd in bdevs:
       stats = rd.GetProcStatus()
 
-      all_connected = (all_connected and
-                       (stats.is_connected or stats.is_in_resync))
+      if multimaster:
+        # In the multimaster case we have to wait explicitly until
+        # the resource is Connected and UpToDate/UpToDate, because
+        # we promote *both nodes* to primary directly afterwards.
+        # Being in resync is not enough, since there is a race during which we
+        # may promote a node with an Outdated disk to primary, effectively
+        # tearing down the connection.
+        all_connected = (all_connected and
+                         stats.is_connected and
+                         stats.is_disk_uptodate and
+                         stats.peer_disk_uptodate)
+      else:
+        all_connected = (all_connected and
+                         (stats.is_connected or stats.is_in_resync))
 
       if stats.is_standalone:
         # peer had different config info and this node became
@@ -3865,7 +3984,7 @@ def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
         _Fail("Can't change to primary mode: %s", err)
 
 
-def DrbdWaitSync(nodes_ip, disks):
+def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
   """Wait until DRBDs have synchronized.
 
   """
@@ -3875,7 +3994,7 @@ def DrbdWaitSync(nodes_ip, disks):
       raise utils.RetryAgain()
     return stats
 
-  bdevs = _FindDisks(nodes_ip, disks)
+  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
 
   min_resync = 100
   alldone = True
@@ -3895,6 +4014,26 @@ def DrbdWaitSync(nodes_ip, disks):
   return (alldone, min_resync)
 
 
+def DrbdNeedsActivation(target_node_uuid, nodes_ip, disks):
+  """Checks which of the passed disks needs activation and returns their UUIDs.
+
+  """
+  _SetPhysicalId(target_node_uuid, nodes_ip, disks)
+  faulty_disks = []
+
+  for disk in disks:
+    rd = _RecursiveFindBD(disk)
+    if rd is None:
+      faulty_disks.append(disk)
+      continue
+
+    stats = rd.GetProcStatus()
+    if stats.is_standalone or stats.is_diskless:
+      faulty_disks.append(disk)
+
+  return [disk.uuid for disk in faulty_disks]
+
+
 def GetDrbdUsermodeHelper():
   """Returns DRBD usermode helper currently configured.
 
@@ -3905,7 +4044,7 @@ def GetDrbdUsermodeHelper():
     _Fail(str(err))
 
 
-def PowercycleNode(hypervisor_type):
+def PowercycleNode(hypervisor_type, hvparams=None):
   """Hard-powercycle the node.
 
   Because we need to return first, and schedule the powercycle in the
@@ -3926,7 +4065,7 @@ def PowercycleNode(hypervisor_type):
   except Exception: # pylint: disable=W0703
     pass
   time.sleep(5)
-  hyper.PowercycleNode()
+  hyper.PowercycleNode(hvparams=hvparams)
 
 
 def _VerifyRestrictedCmdName(cmd):