import random
import logging
import tempfile
+import zlib
+import base64
from ganeti import errors
from ganeti import utils
from ganeti import ssconf
+class RPCFail(Exception):
+ """Class denoting RPC failure.
+
+ Its argument is the error message.
+
+ """
+
+def _Fail(msg, *args, **kwargs):
+ """Log an error and the raise an RPCFail exception.
+
+ This exception is then handled specially in the ganeti daemon and
+ turned into a 'failed' return type. As such, this function is a
+ useful shortcut for logging the error and returning it to the master
+ daemon.
+
+ @type msg: string
+ @param msg: the text of the exception
+ @raise RPCFail
+
+ """
+ if args:
+ msg = msg % args
+ if "exc" in kwargs and kwargs["exc"]:
+ logging.exception(msg)
+ else:
+ logging.error(msg)
+ raise RPCFail(msg)
+
+
def _GetConfig():
- """Simple wrapper to return a ConfigReader.
+ """Simple wrapper to return a SimpleStore.
- @rtype: L{ssconf.SimpleConfigReader}
- @return: a SimpleConfigReader instance
+ @rtype: L{ssconf.SimpleStore}
+ @return: a SimpleStore instance
"""
- return ssconf.SimpleConfigReader()
+ return ssconf.SimpleStore()
def _GetSshRunner(cluster_name):
return ssh.SshRunner(cluster_name)
-def _CleanDirectory(path, exclude=[]):
+def _Decompress(data):
+ """Unpacks data compressed by the RPC client.
+
+ @type data: list or tuple
+ @param data: Data sent by RPC client
+ @rtype: str
+ @return: Decompressed data
+
+ """
+ assert isinstance(data, (list, tuple))
+ assert len(data) == 2
+ (encoding, content) = data
+ if encoding == constants.RPC_ENCODING_NONE:
+ return content
+ elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
+ return zlib.decompress(base64.b64decode(content))
+ else:
+ raise AssertionError("Unknown data encoding")
+
+
+def _CleanDirectory(path, exclude=None):
"""Removes all regular files in a directory.
@type path: str
@type exclude: list
@param exclude: list of files to be excluded, defaults
to the empty list
- @rtype: None
"""
if not os.path.isdir(path):
return
-
- # Normalize excluded paths
- exclude = [os.path.normpath(i) for i in exclude]
+ if exclude is None:
+ exclude = []
+ else:
+ # Normalize excluded paths
+ exclude = [os.path.normpath(i) for i in exclude]
for rel_name in utils.ListVisibleFiles(path):
full_name = os.path.normpath(os.path.join(path, rel_name))
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
mkdir=True)
except errors.OpExecError, err:
- logging.exception("Error while processing user ssh files")
- return False
+ _Fail("Error while processing user ssh files: %s", err, exc=True)
for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
utils.WriteFile(name, data=content, mode=0600)
utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
- return True
+ return (True, "Node added successfully")
def LeaveCluster():
from the cluster.
If processing is successful, then it raises an
- L{errors.GanetiQuitException} which is used as a special case to
+ L{errors.QuitGanetiException} which is used as a special case to
shutdown the node daemon.
"""
"""
result = {}
- if 'hypervisor' in what:
- result['hypervisor'] = my_dict = {}
- for hv_name in what['hypervisor']:
- my_dict[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
+ if constants.NV_HYPERVISOR in what:
+ result[constants.NV_HYPERVISOR] = tmp = {}
+ for hv_name in what[constants.NV_HYPERVISOR]:
+ tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
- if 'filelist' in what:
- result['filelist'] = utils.FingerprintFiles(what['filelist'])
+ if constants.NV_FILELIST in what:
+ result[constants.NV_FILELIST] = utils.FingerprintFiles(
+ what[constants.NV_FILELIST])
- if 'nodelist' in what:
- result['nodelist'] = {}
- random.shuffle(what['nodelist'])
- for node in what['nodelist']:
+ if constants.NV_NODELIST in what:
+ result[constants.NV_NODELIST] = tmp = {}
+ random.shuffle(what[constants.NV_NODELIST])
+ for node in what[constants.NV_NODELIST]:
success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
if not success:
- result['nodelist'][node] = message
- if 'node-net-test' in what:
- result['node-net-test'] = {}
+ tmp[node] = message
+
+ if constants.NV_NODENETTEST in what:
+ result[constants.NV_NODENETTEST] = tmp = {}
my_name = utils.HostInfo().name
my_pip = my_sip = None
- for name, pip, sip in what['node-net-test']:
+ for name, pip, sip in what[constants.NV_NODENETTEST]:
if name == my_name:
my_pip = pip
my_sip = sip
break
if not my_pip:
- result['node-net-test'][my_name] = ("Can't find my own"
- " primary/secondary IP"
- " in the node list")
+ tmp[my_name] = ("Can't find my own primary/secondary IP"
+ " in the node list")
else:
port = utils.GetNodeDaemonPort()
- for name, pip, sip in what['node-net-test']:
+ for name, pip, sip in what[constants.NV_NODENETTEST]:
fail = []
if not utils.TcpPing(pip, port, source=my_pip):
fail.append("primary")
if not utils.TcpPing(sip, port, source=my_sip):
fail.append("secondary")
if fail:
- result['node-net-test'][name] = ("failure using the %s"
- " interface(s)" %
- " and ".join(fail))
+ tmp[name] = ("failure using the %s interface(s)" %
+ " and ".join(fail))
+
+ if constants.NV_LVLIST in what:
+ result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
+
+ if constants.NV_INSTANCELIST in what:
+ result[constants.NV_INSTANCELIST] = GetInstanceList(
+ what[constants.NV_INSTANCELIST])
+
+ if constants.NV_VGLIST in what:
+ result[constants.NV_VGLIST] = ListVolumeGroups()
+
+ if constants.NV_VERSION in what:
+ result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
+ constants.RELEASE_VERSION)
+
+ if constants.NV_HVINFO in what:
+ hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
+ result[constants.NV_HVINFO] = hyper.GetNodeInfo()
+
+ if constants.NV_DRBDLIST in what:
+ try:
+ used_minors = bdev.DRBD8.GetUsedDevs().keys()
+ except errors.BlockDeviceError, err:
+ logging.warning("Can't get used minors list", exc_info=True)
+ used_minors = str(err)
+ result[constants.NV_DRBDLIST] = used_minors
return result
results.extend(names)
except errors.HypervisorError, err:
logging.exception("Error enumerating instances for hypevisor %s", hname)
- # FIXME: should we somehow not propagate this to the master?
raise
return results
return output
+def GetInstanceMigratable(instance):
+ """Gives whether an instance can be migrated.
+
+ @type instance: L{objects.Instance}
+ @param instance: object representing the instance to be checked.
+
+ @rtype: tuple
+ @return: tuple of (result, description) where:
+ - result: whether the instance can be migrated or not
+ - description: a description of the issue, if relevant
+
+ """
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ if instance.name not in hyper.ListInstances():
+ return (False, 'not running')
+
+ for idx in range(len(instance.disks)):
+ link_name = _GetBlockDevSymlinkPath(instance.name, idx)
+ if not os.path.islink(link_name):
+ return (False, 'not restarted since ganeti 1.2.5')
+
+ return (True, '')
+
+
def GetAllInstancesInfo(hypervisor_list):
"""Gather data about all instances.
@type hypervisor_list: list
@param hypervisor_list: list of hypervisors to query for instance data
- @rtype: dict of dicts
+ @rtype: dict
@return: dictionary of instance: data, with data having the following keys:
- memory: memory size of instance (int)
- state: xen state of instance (string)
'state': state,
'time': times,
}
- if name in output and output[name] != value:
- raise errors.HypervisorError("Instance %s running duplicate"
- " with different parameters" % name)
+ if name in output:
+ # we only check static parameters, like memory and vcpus,
+ # and not state and time which can change between the
+ # invocations of the different hypervisors
+ for key in 'memory', 'vcpus':
+ if value[key] != output[name][key]:
+ raise errors.HypervisorError("Instance %s is running twice"
+ " with different parameters" % name)
output[name] = value
return output
-def AddOSToInstance(instance):
+def InstanceOsAdd(instance, reinstall):
"""Add an OS to an instance.
@type instance: L{objects.Instance}
@param instance: Instance whose OS is to be installed
+ @type reinstall: boolean
+ @param reinstall: whether this is an instance reinstall
@rtype: boolean
@return: the success of the operation
"""
- inst_os = OSFromDisk(instance.os)
+ try:
+ inst_os = OSFromDisk(instance.os)
+ except errors.InvalidOS, err:
+ os_name, os_dir, os_err = err.args
+ if os_dir is None:
+ return (False, "Can't find OS '%s': %s" % (os_name, os_err))
+ else:
+ return (False, "Error parsing OS '%s' in directory %s: %s" %
+ (os_name, os_dir, os_err))
create_env = OSEnvironment(instance)
+ if reinstall:
+ create_env['INSTANCE_REINSTALL'] = "1"
logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
instance.name, int(time.time()))
logging.error("os create command '%s' returned error: %s, logfile: %s,"
" output: %s", result.cmd, result.fail_reason, logfile,
result.output)
- return False
+ lines = [utils.SafeEncode(val)
+ for val in utils.TailFile(logfile, lines=20)]
+ return (False, "OS create script failed (%s), last lines in the"
+ " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
- return True
+ return (True, "Successfully installed")
def RunRenameInstance(instance, old_name):
if result.failed:
logging.error("os create command '%s' returned error: %s output: %s",
result.cmd, result.fail_reason, result.output)
- return False
+ lines = [utils.SafeEncode(val)
+ for val in utils.TailFile(logfile, lines=20)]
+ return (False, "OS rename script failed (%s), last lines in the"
+ " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
- return True
+ return (True, "Rename successful")
def _GetVGInfo(vg_name):
return retdic
-def _GatherBlockDevs(instance):
+def _GetBlockDevSymlinkPath(instance_name, idx):
+ return os.path.join(constants.DISK_LINKS_DIR,
+ "%s:%d" % (instance_name, idx))
+
+
+def _SymlinkBlockDev(instance_name, device_path, idx):
+ """Set up symlinks to a instance's block device.
+
+ This is an auxiliary function run when an instance is start (on the primary
+ node) or when an instance is migrated (on the target node).
+
+
+ @param instance_name: the name of the target instance
+ @param device_path: path of the physical block device, on the node
+ @param idx: the disk index
+ @return: absolute path to the disk's symlink
+
+ """
+ link_name = _GetBlockDevSymlinkPath(instance_name, idx)
+ try:
+ os.symlink(device_path, link_name)
+ except OSError, err:
+ if err.errno == errno.EEXIST:
+ if (not os.path.islink(link_name) or
+ os.readlink(link_name) != device_path):
+ os.remove(link_name)
+ os.symlink(device_path, link_name)
+ else:
+ raise
+
+ return link_name
+
+
+def _RemoveBlockDevLinks(instance_name, disks):
+ """Remove the block device symlinks belonging to the given instance.
+
+ """
+ for idx, disk in enumerate(disks):
+ link_name = _GetBlockDevSymlinkPath(instance_name, idx)
+ if os.path.islink(link_name):
+ try:
+ os.remove(link_name)
+ except OSError:
+ logging.exception("Can't remove symlink '%s'", link_name)
+
+
+def _GatherAndLinkBlockDevs(instance):
"""Set up an instance's block device(s).
This is run on the primary node at instance startup. The block
@type instance: L{objects.Instance}
@param instance: the instance whose disks we shoul assemble
- @rtype: list of L{bdev.BlockDev}
- @return: list of the block devices
+ @rtype: list
+ @return: list of (disk_object, device_path)
"""
block_devices = []
- for disk in instance.disks:
+ for idx, disk in enumerate(instance.disks):
device = _RecursiveFindBD(disk)
if device is None:
raise errors.BlockDeviceError("Block device '%s' is not set up." %
str(disk))
device.Open()
- block_devices.append((disk, device))
+ try:
+ link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
+ except OSError, e:
+ raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
+ e.strerror)
+
+ block_devices.append((disk, link_name))
+
return block_devices
-def StartInstance(instance, extra_args):
+def StartInstance(instance):
"""Start an instance.
@type instance: L{objects.Instance}
running_instances = GetInstanceList([instance.hypervisor])
if instance.name in running_instances:
- return True
-
- block_devices = _GatherBlockDevs(instance)
- hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ return (True, "Already running")
try:
- hyper.StartInstance(instance, block_devices, extra_args)
+ block_devices = _GatherAndLinkBlockDevs(instance)
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ hyper.StartInstance(instance, block_devices)
+ except errors.BlockDeviceError, err:
+ _Fail("Block device error: %s", err, exc=True)
except errors.HypervisorError, err:
- logging.exception("Failed to start instance")
- return False
+ _RemoveBlockDevLinks(instance.name, instance.disks)
+ _Fail("Hypervisor error: %s", err, exc=True)
- return True
+ return (True, "Instance started successfully")
-def ShutdownInstance(instance):
+def InstanceShutdown(instance):
"""Shut an instance down.
@note: this functions uses polling with a hardcoded timeout.
running_instances = GetInstanceList([hv_name])
if instance.name not in running_instances:
- return True
+ return (True, "Instance already stopped")
hyper = hypervisor.GetHypervisor(hv_name)
try:
hyper.StopInstance(instance)
except errors.HypervisorError, err:
- logging.error("Failed to stop instance")
- return False
+ _Fail("Failed to stop instance %s: %s", instance.name, err)
# test every 10secs for 2min
- shutdown_ok = False
time.sleep(1)
for dummy in range(11):
time.sleep(10)
else:
# the shutdown did not succeed
- logging.error("shutdown of '%s' unsuccessful, using destroy", instance)
+ logging.error("Shutdown of '%s' unsuccessful, using destroy",
+ instance.name)
try:
hyper.StopInstance(instance, force=True)
except errors.HypervisorError, err:
- logging.exception("Failed to stop instance")
- return False
+ _Fail("Failed to force stop instance %s: %s", instance.name, err)
time.sleep(1)
if instance.name in GetInstanceList([hv_name]):
- logging.error("could not shutdown instance '%s' even by destroy",
- instance.name)
- return False
+ _Fail("Could not shutdown instance %s even by destroy", instance.name)
- return True
+ _RemoveBlockDevLinks(instance.name, instance.disks)
+
+ return (True, "Instance has been shutdown successfully")
-def RebootInstance(instance, reboot_type, extra_args):
+def InstanceReboot(instance, reboot_type):
"""Reboot an instance.
@type instance: L{objects.Instance}
running_instances = GetInstanceList([instance.hypervisor])
if instance.name not in running_instances:
- logging.error("Cannot reboot instance that is not running")
- return False
+ _Fail("Cannot reboot instance %s that is not running", instance.name)
hyper = hypervisor.GetHypervisor(instance.hypervisor)
if reboot_type == constants.INSTANCE_REBOOT_SOFT:
try:
hyper.RebootInstance(instance)
except errors.HypervisorError, err:
- logging.exception("Failed to soft reboot instance")
- return False
+ _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
elif reboot_type == constants.INSTANCE_REBOOT_HARD:
try:
- ShutdownInstance(instance)
- StartInstance(instance, extra_args)
+ stop_result = InstanceShutdown(instance)
+ if not stop_result[0]:
+ return stop_result
+ return StartInstance(instance)
except errors.HypervisorError, err:
- logging.exception("Failed to hard reboot instance")
- return False
+ _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
else:
- raise errors.ParameterError("reboot_type invalid")
+ _Fail("Invalid reboot_type received: %s", reboot_type)
+
+ return (True, "Reboot successful")
- return True
+
+def MigrationInfo(instance):
+ """Gather information about an instance to be migrated.
+
+ @type instance: L{objects.Instance}
+ @param instance: the instance definition
+
+ """
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ try:
+ info = hyper.MigrationInfo(instance)
+ except errors.HypervisorError, err:
+ _Fail("Failed to fetch migration information: %s", err, exc=True)
+ return (True, info)
+
+
+def AcceptInstance(instance, info, target):
+ """Prepare the node to accept an instance.
+
+ @type instance: L{objects.Instance}
+ @param instance: the instance definition
+ @type info: string/data (opaque)
+ @param info: migration information, from the source node
+ @type target: string
+ @param target: target host (usually ip), on this node
+
+ """
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ try:
+ hyper.AcceptInstance(instance, info, target)
+ except errors.HypervisorError, err:
+ _Fail("Failed to accept instance: %s", err, exc=True)
+ return (True, "Accept successfull")
+
+
+def FinalizeMigration(instance, info, success):
+ """Finalize any preparation to accept an instance.
+
+ @type instance: L{objects.Instance}
+ @param instance: the instance definition
+ @type info: string/data (opaque)
+ @param info: migration information, from the source node
+ @type success: boolean
+ @param success: whether the migration was a success or a failure
+
+ """
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
+ try:
+ hyper.FinalizeMigration(instance, info, success)
+ except errors.HypervisorError, err:
+ _Fail("Failed to finalize migration: %s", err, exc=True)
+ return (True, "Migration Finalized")
def MigrateInstance(instance, target, live):
- msg is a string with details in case of failure
"""
- hyper = hypervisor.GetHypervisor(instance.hypervisor_name)
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
try:
hyper.MigrateInstance(instance.name, target, live)
except errors.HypervisorError, err:
- msg = "Failed to migrate instance: %s" % str(err)
- logging.error(msg)
- return (False, msg)
+ _Fail("Failed to migrate instance: %s", err, exc=True)
return (True, "Migration successfull")
-def CreateBlockDevice(disk, size, owner, on_primary, info):
+def BlockdevCreate(disk, size, owner, on_primary, info):
"""Creates a block device for an instance.
@type disk: L{objects.Disk}
clist = []
if disk.children:
for child in disk.children:
- crdev = _RecursiveAssembleBD(child, owner, on_primary)
+ try:
+ crdev = _RecursiveAssembleBD(child, owner, on_primary)
+ except errors.BlockDeviceError, err:
+ _Fail("Can't assemble device %s: %s", child, err)
if on_primary or disk.AssembleOnSecondary():
# we need the children open in case the device itself has to
# be assembled
- crdev.Open()
+ try:
+ crdev.Open()
+ except errors.BlockDeviceError, err:
+ _Fail("Can't make child '%s' read-write: %s", child, err)
clist.append(crdev)
+
try:
- device = bdev.FindDevice(disk.dev_type, disk.physical_id, clist)
- if device is not None:
- logging.info("removing existing device %s", disk)
- device.Remove()
+ device = bdev.Create(disk.dev_type, disk.physical_id, clist, size)
except errors.BlockDeviceError, err:
- pass
+ _Fail("Can't create block device: %s", err)
- device = bdev.Create(disk.dev_type, disk.physical_id,
- clist, size)
- if device is None:
- raise ValueError("Can't create child device for %s, %s" %
- (disk, size))
if on_primary or disk.AssembleOnSecondary():
- if not device.Assemble():
- errorstring = "Can't assemble device after creation"
- logging.error(errorstring)
- raise errors.BlockDeviceError("%s, very unusual event - check the node"
- " daemon logs" % errorstring)
+ try:
+ device.Assemble()
+ except errors.BlockDeviceError, err:
+ _Fail("Can't assemble device after creation, unusual event: %s", err)
device.SetSyncSpeed(constants.SYNC_SPEED)
if on_primary or disk.OpenOnSecondary():
- device.Open(force=True)
+ try:
+ device.Open(force=True)
+ except errors.BlockDeviceError, err:
+ _Fail("Can't make device r/w after creation, unusual event: %s", err)
DevCacheManager.UpdateCache(device.dev_path, owner,
on_primary, disk.iv_name)
device.SetInfo(info)
physical_id = device.unique_id
- return physical_id
+ return True, physical_id
-def RemoveBlockDevice(disk):
+def BlockdevRemove(disk):
"""Remove a block device.
@note: This is intended to be called recursively.
- @type disk: L{objects.disk}
+ @type disk: L{objects.Disk}
@param disk: the disk object we should remove
@rtype: boolean
@return: the success of the operation
"""
+ msgs = []
+ result = True
try:
- # since we are removing the device, allow a partial match
- # this allows removal of broken mirrors
- rdev = _RecursiveFindBD(disk, allow_partial=True)
+ rdev = _RecursiveFindBD(disk)
except errors.BlockDeviceError, err:
# probably can't attach
logging.info("Can't attach to device %s in remove", disk)
rdev = None
if rdev is not None:
r_path = rdev.dev_path
- result = rdev.Remove()
+ try:
+ rdev.Remove()
+ except errors.BlockDeviceError, err:
+ msgs.append(str(err))
+ result = False
if result:
DevCacheManager.RemoveCache(r_path)
- else:
- result = True
+
if disk.children:
for child in disk.children:
- result = result and RemoveBlockDevice(child)
- return result
+ c_status, c_msg = BlockdevRemove(child)
+ result = result and c_status
+ if c_msg: # not an empty message
+ msgs.append(c_msg)
+
+ return (result, "; ".join(msgs))
def _RecursiveAssembleBD(disk, owner, as_primary):
if children.count(None) >= mcn:
raise
cdev = None
- logging.debug("Error in child activation: %s", str(err))
+ logging.error("Error in child activation (but continuing): %s",
+ str(err))
children.append(cdev)
if as_primary or disk.AssembleOnSecondary():
- r_dev = bdev.AttachOrAssemble(disk.dev_type, disk.physical_id, children)
+ r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children)
r_dev.SetSyncSpeed(constants.SYNC_SPEED)
result = r_dev
if as_primary or disk.OpenOnSecondary():
return result
-def AssembleBlockDevice(disk, owner, as_primary):
+def BlockdevAssemble(disk, owner, as_primary):
"""Activate a block device for an instance.
This is a wrapper over _RecursiveAssembleBD.
C{True} for secondary nodes
"""
- result = _RecursiveAssembleBD(disk, owner, as_primary)
- if isinstance(result, bdev.BlockDev):
- result = result.dev_path
- return result
+ status = True
+ result = "no error information"
+ try:
+ result = _RecursiveAssembleBD(disk, owner, as_primary)
+ if isinstance(result, bdev.BlockDev):
+ result = result.dev_path
+ except errors.BlockDeviceError, err:
+ result = "Error while assembling disk: %s" % str(err)
+ status = False
+ return (status, result)
-def ShutdownBlockDevice(disk):
+def BlockdevShutdown(disk):
"""Shut down a block device.
- First, if the device is assembled (can L{Attach()}), then the device
- is shutdown. Then the children of the device are shutdown.
+ First, if the device is assembled (Attach() is successfull), then
+ the device is shutdown. Then the children of the device are
+ shutdown.
This function is called recursively. Note that we don't cache the
children or such, as oppossed to assemble, shutdown of different
@return: the success of the operation
"""
+ msgs = []
+ result = True
r_dev = _RecursiveFindBD(disk)
if r_dev is not None:
r_path = r_dev.dev_path
- result = r_dev.Shutdown()
- if result:
+ try:
+ r_dev.Shutdown()
DevCacheManager.RemoveCache(r_path)
- else:
- result = True
+ except errors.BlockDeviceError, err:
+ msgs.append(str(err))
+ result = False
+
if disk.children:
for child in disk.children:
- result = result and ShutdownBlockDevice(child)
- return result
+ c_status, c_msg = BlockdevShutdown(child)
+ result = result and c_status
+ if c_msg: # not an empty message
+ msgs.append(c_msg)
+ return (result, "; ".join(msgs))
-def MirrorAddChildren(parent_cdev, new_cdevs):
+
+def BlockdevAddchildren(parent_cdev, new_cdevs):
"""Extend a mirrored block device.
@type parent_cdev: L{objects.Disk}
@return: the success of the operation
"""
- parent_bdev = _RecursiveFindBD(parent_cdev, allow_partial=True)
+ parent_bdev = _RecursiveFindBD(parent_cdev)
if parent_bdev is None:
- logging.error("Can't find parent device")
- return False
+ _Fail("Can't find parent device '%s' in add children", parent_cdev)
new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
if new_bdevs.count(None) > 0:
- logging.error("Can't find new device(s) to add: %s:%s",
- new_bdevs, new_cdevs)
- return False
+ _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
parent_bdev.AddChildren(new_bdevs)
- return True
+ return (True, None)
-def MirrorRemoveChildren(parent_cdev, new_cdevs):
+def BlockdevRemovechildren(parent_cdev, new_cdevs):
"""Shrink a mirrored block device.
@type parent_cdev: L{objects.Disk}
"""
parent_bdev = _RecursiveFindBD(parent_cdev)
if parent_bdev is None:
- logging.error("Can't find parent in remove children: %s", parent_cdev)
- return False
+ _Fail("Can't find parent device '%s' in remove children", parent_cdev)
devs = []
for disk in new_cdevs:
rpath = disk.StaticDevPath()
if rpath is None:
bd = _RecursiveFindBD(disk)
if bd is None:
- logging.error("Can't find dynamic device %s while removing children",
- disk)
- return False
+ _Fail("Can't find device %s while removing children", disk)
else:
devs.append(bd.dev_path)
else:
devs.append(rpath)
parent_bdev.RemoveChildren(devs)
- return True
+ return (True, None)
-def GetMirrorStatus(disks):
+def BlockdevGetmirrorstatus(disks):
"""Get the mirroring status of a list of devices.
@type disks: list of L{objects.Disk}
@rtype: disk
@return:
a list of (mirror_done, estimated_time) tuples, which
- are the result of L{bdev.BlockDevice.CombinedSyncStatus}
+ are the result of L{bdev.BlockDev.CombinedSyncStatus}
@raise errors.BlockDeviceError: if any of the disks cannot be
found
for dsk in disks:
rbd = _RecursiveFindBD(dsk)
if rbd is None:
- raise errors.BlockDeviceError("Can't find device %s" % str(dsk))
+ _Fail("Can't find device %s", dsk)
stats.append(rbd.CombinedSyncStatus())
- return stats
+ return True, stats
-def _RecursiveFindBD(disk, allow_partial=False):
+def _RecursiveFindBD(disk):
"""Check if a device is activated.
If so, return informations about the real device.
@type disk: L{objects.Disk}
@param disk: the disk object we need to find
- @type allow_partial: boolean
- @param allow_partial: if true, don't abort the find if a
- child of the device can't be found; this is intended
- to be used when repairing mirrors
@return: None if the device can't be found,
otherwise the device instance
return bdev.FindDevice(disk.dev_type, disk.physical_id, children)
-def FindBlockDevice(disk):
+def BlockdevFind(disk):
"""Check if a device is activated.
If it is, return informations about the real device.
estimated_time, is_degraded)
"""
- rbd = _RecursiveFindBD(disk)
+ try:
+ rbd = _RecursiveFindBD(disk)
+ except errors.BlockDeviceError, err:
+ _Fail("Failed to find device: %s", err, exc=True)
if rbd is None:
- return rbd
- return (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus()
+ return (True, None)
+ return (True, (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus())
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
"""
if not os.path.isabs(file_name):
- logging.error("Filename passed to UploadFile is not absolute: '%s'",
- file_name)
- return False
+ _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
- allowed_files = [
+ allowed_files = set([
constants.CLUSTER_CONF_FILE,
constants.ETC_HOSTS,
constants.SSH_KNOWN_HOSTS_FILE,
constants.VNC_PASSWORD_FILE,
- ]
+ constants.RAPI_CERT_FILE,
+ constants.RAPI_USERS_FILE,
+ ])
+
+ for hv_name in constants.HYPER_TYPES:
+ hv_class = hypervisor.GetHypervisor(hv_name)
+ allowed_files.update(hv_class.GetAncillaryFiles())
if file_name not in allowed_files:
- logging.error("Filename passed to UploadFile not in allowed"
- " upload targets: '%s'", file_name)
- return False
+ _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
+ file_name)
- utils.WriteFile(file_name, data=data, mode=mode, uid=uid, gid=gid,
+ raw_data = _Decompress(data)
+
+ utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
atime=atime, mtime=mtime)
- return True
+ return (True, "success")
+
+def WriteSsconfFiles(values):
+ """Update all ssconf files.
-def WriteSsconfFiles():
- ssconf.WriteSsconfFiles(constants.CLUSTER_CONF_FILE)
+ Wrapper around the SimpleStore.WriteFiles.
+
+ """
+ ssconf.SimpleStore().WriteFiles(values)
def _ErrnoOrStr(err):
result = {}
result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
result['INSTANCE_NAME'] = instance.name
+ result['INSTANCE_OS'] = instance.os
result['HYPERVISOR'] = instance.hypervisor
result['DISK_COUNT'] = '%d' % len(instance.disks)
result['NIC_COUNT'] = '%d' % len(instance.nics)
str(disk))
real_disk.Open()
result['DISK_%d_PATH' % idx] = real_disk.dev_path
- # FIXME: When disks will have read-only mode, populate this
- result['DISK_%d_ACCESS' % idx] = 'W'
+ result['DISK_%d_ACCESS' % idx] = disk.mode
if constants.HV_DISK_TYPE in instance.hvparams:
result['DISK_%d_FRONTEND_TYPE' % idx] = \
instance.hvparams[constants.HV_DISK_TYPE]
result['NIC_%d_MAC' % idx] = nic.mac
if nic.ip:
result['NIC_%d_IP' % idx] = nic.ip
- result['NIC_%d_BRIDGE' % idx] = nic.bridge
+ result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
+ if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
+ result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
+ if nic.nicparams[constants.NIC_LINK]:
+ result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
if constants.HV_NIC_TYPE in instance.hvparams:
result['NIC_%d_FRONTEND_TYPE' % idx] = \
instance.hvparams[constants.HV_NIC_TYPE]
return result
-def GrowBlockDevice(disk, amount):
+def BlockdevGrow(disk, amount):
"""Grow a stack of block devices.
This function is called recursively, with the childrens being the
try:
r_dev.Grow(amount)
except errors.BlockDeviceError, err:
- return False, str(err)
+ _Fail("Failed to grow block device: %s", err, exc=True)
return True, None
-def SnapshotBlockDevice(disk):
+def BlockdevSnapshot(disk):
"""Create a snapshot copy of a block device.
This function is called recursively, and the snapshot is actually created
if disk.children:
if len(disk.children) == 1:
# only one child, let's recurse on it
- return SnapshotBlockDevice(disk.children[0])
+ return BlockdevSnapshot(disk.children[0])
else:
# more than one child, choose one that matches
for child in disk.children:
if child.size == disk.size:
# return implies breaking the loop
- return SnapshotBlockDevice(child)
+ return BlockdevSnapshot(child)
elif disk.dev_type == constants.LD_LV:
r_dev = _RecursiveFindBD(disk)
if r_dev is not None:
# let's stay on the safe side and ask for the full size, for now
- return r_dev.Snapshot(disk.size)
+ return True, r_dev.Snapshot(disk.size)
else:
- return None
+ _Fail("Cannot find block device %s", disk)
else:
- raise errors.ProgrammerError("Cannot snapshot non-lvm block device"
- " '%s' of type '%s'" %
- (disk.unique_id, disk.dev_type))
+ _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
+ disk.unique_id, disk.dev_type)
def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
os.mkdir(constants.LOG_OS_DIR, 0750)
real_disk = _RecursiveFindBD(disk)
if real_disk is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(disk))
+ _Fail("Block device '%s' is not set up", disk)
+
real_disk.Open()
export_env['EXPORT_DEVICE'] = real_disk.dev_path
result = utils.RunCmd(command, env=export_env)
if result.failed:
- logging.error("os snapshot export command '%s' returned error: %s"
- " output: %s", command, result.fail_reason, result.output)
- return False
+ _Fail("OS snapshot export command '%s' returned error: %s"
+ " output: %s", command, result.fail_reason, result.output)
- return True
+ return (True, None)
def FinalizeExport(instance, snap_disks):
instance.beparams[constants.BE_VCPUS])
config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
- nic_count = 0
+ nic_total = 0
for nic_count, nic in enumerate(instance.nics):
+ nic_total += 1
config.set(constants.INISECT_INS, 'nic%d_mac' %
nic_count, '%s' % nic.mac)
config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
'%s' % nic.bridge)
# TODO: redundant: on load can read nics until it doesn't exist
- config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_count)
+ config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
- disk_count = 0
+ disk_total = 0
for disk_count, disk in enumerate(snap_disks):
if disk:
+ disk_total += 1
config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
('%s' % disk.iv_name))
config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
('%s' % disk.physical_id[1]))
config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
('%d' % disk.size))
- config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_count)
- cff = os.path.join(destdir, constants.EXPORT_CONF_FILE)
- cfo = open(cff, 'w')
- try:
- config.write(cfo)
- finally:
- cfo.close()
+ config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
+ utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
+ data=config.Dumps())
shutil.rmtree(finaldestdir, True)
shutil.move(destdir, finaldestdir)
import_env['IMPORT_INDEX'] = str(idx)
result = utils.RunCmd(command, env=import_env)
if result.failed:
- logging.error("disk import command '%s' returned error: %s"
- " output: %s", command, result.fail_reason, result.output)
+ logging.error("Disk import command '%s' returned error: %s"
+ " output: %s", command, result.fail_reason,
+ result.output)
final_result.append(False)
else:
final_result.append(True)
return True
-def RenameBlockDevices(devlist):
+def BlockdevRename(devlist):
"""Rename a list of block devices.
@type devlist: list of tuples
@return: True if all renames succeeded, False otherwise
"""
+ msgs = []
result = True
for disk, unique_id in devlist:
dev = _RecursiveFindBD(disk)
if dev is None:
+ msgs.append("Can't find device %s in rename" % str(disk))
result = False
continue
try:
# cache? for now, we only lose lvm data when we rename, which
# is less critical than DRBD or MD
except errors.BlockDeviceError, err:
+ msgs.append("Can't rename device '%s' to '%s': %s" %
+ (dev, unique_id, err))
logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
result = False
- return result
+ return (result, "; ".join(msgs))
def _TransformFileStorageDir(file_storage_dir):
return False
# Write and replace the file atomically
- utils.WriteFile(file_name, data=content)
+ utils.WriteFile(file_name, data=_Decompress(content))
return True
def JobQueueRename(old, new):
"""Renames a job queue file.
- This is just a wrapper over L{os.rename} with proper checking.
+ This is just a wrapper over os.rename with proper checking.
@type old: str
@param old: the old (actual) file name
if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)):
return False
- os.rename(old, new)
+ utils.RenameFile(old, new, mkdir=True)
return True
return True
-def CloseBlockDevices(disks):
+def BlockdevClose(instance_name, disks):
"""Closes the given block devices.
This means they will be switched to secondary mode (in case of
DRBD).
+ @param instance_name: if the argument is not empty, the symlinks
+ of this instance will be removed
@type disks: list of L{objects.Disk}
@param disks: the list of disks to be closed
@rtype: tuple (success, message)
for cf in disks:
rd = _RecursiveFindBD(cf)
if rd is None:
- return (False, "Can't find device %s" % cf)
+ _Fail("Can't find device %s", cf)
bdevs.append(rd)
msg = []
if msg:
return (False, "Can't make devices secondary: %s" % ",".join(msg))
else:
+ if instance_name:
+ _RemoveBlockDevLinks(instance_name, disks)
return (True, "All devices secondary")
return (False, str(err))
+def DemoteFromMC():
+ """Demotes the current node from master candidate role.
+
+ """
+ # try to ensure we're not the master by mistake
+ master, myself = ssconf.GetMasterAndMyself()
+ if master == myself:
+ return (False, "ssconf status shows I'm the master node, will not demote")
+ pid_file = utils.DaemonPidFileName(constants.MASTERD_PID)
+ if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
+ return (False, "The master daemon is running, will not demote")
+ try:
+ utils.CreateBackup(constants.CLUSTER_CONF_FILE)
+ except EnvironmentError, err:
+ if err.errno != errno.ENOENT:
+ return (False, "Error while backing up cluster file: %s" % str(err))
+ utils.RemoveFile(constants.CLUSTER_CONF_FILE)
+ return (True, "Done")
+
+
+def _FindDisks(nodes_ip, disks):
+ """Sets the physical ID on disks and returns the block devices.
+
+ """
+ # set the correct physical ID
+ my_name = utils.HostInfo().name
+ for cf in disks:
+ cf.SetPhysicalID(my_name, nodes_ip)
+
+ bdevs = []
+
+ for cf in disks:
+ rd = _RecursiveFindBD(cf)
+ if rd is None:
+ return (False, "Can't find device %s" % cf)
+ bdevs.append(rd)
+ return (True, bdevs)
+
+
+def DrbdDisconnectNet(nodes_ip, disks):
+ """Disconnects the network on a list of drbd devices.
+
+ """
+ status, bdevs = _FindDisks(nodes_ip, disks)
+ if not status:
+ return status, bdevs
+
+ # disconnect disks
+ for rd in bdevs:
+ try:
+ rd.DisconnectNet()
+ except errors.BlockDeviceError, err:
+ _Fail("Can't change network configuration to standalone mode: %s",
+ err, exc=True)
+ return (True, "All disks are now disconnected")
+
+
+def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
+ """Attaches the network on a list of drbd devices.
+
+ """
+ status, bdevs = _FindDisks(nodes_ip, disks)
+ if not status:
+ return status, bdevs
+
+ if multimaster:
+ for idx, rd in enumerate(bdevs):
+ try:
+ _SymlinkBlockDev(instance_name, rd.dev_path, idx)
+ except EnvironmentError, err:
+ _Fail("Can't create symlink: %s", err)
+ # reconnect disks, switch to new master configuration and if
+ # needed primary mode
+ for rd in bdevs:
+ try:
+ rd.AttachNet(multimaster)
+ except errors.BlockDeviceError, err:
+ _Fail("Can't change network configuration: %s", err)
+ # wait until the disks are connected; we need to retry the re-attach
+ # if the device becomes standalone, as this might happen if the one
+ # node disconnects and reconnects in a different mode before the
+ # other node reconnects; in this case, one or both of the nodes will
+ # decide it has wrong configuration and switch to standalone
+ RECONNECT_TIMEOUT = 2 * 60
+ sleep_time = 0.100 # start with 100 miliseconds
+ timeout_limit = time.time() + RECONNECT_TIMEOUT
+ while time.time() < timeout_limit:
+ all_connected = True
+ for rd in bdevs:
+ stats = rd.GetProcStatus()
+ if not (stats.is_connected or stats.is_in_resync):
+ all_connected = False
+ if stats.is_standalone:
+ # peer had different config info and this node became
+ # standalone, even though this should not happen with the
+ # new staged way of changing disk configs
+ try:
+ rd.ReAttachNet(multimaster)
+ except errors.BlockDeviceError, err:
+ _Fail("Can't change network configuration: %s", err)
+ if all_connected:
+ break
+ time.sleep(sleep_time)
+ sleep_time = min(5, sleep_time * 1.5)
+ if not all_connected:
+ return (False, "Timeout in disk reconnecting")
+ if multimaster:
+ # change to primary mode
+ for rd in bdevs:
+ try:
+ rd.Open()
+ except errors.BlockDeviceError, err:
+ _Fail("Can't change to primary mode: %s", err)
+ if multimaster:
+ msg = "multi-master and primary"
+ else:
+ msg = "single-master"
+ return (True, "Disks are now configured as %s" % msg)
+
+
+def DrbdWaitSync(nodes_ip, disks):
+ """Wait until DRBDs have synchronized.
+
+ """
+ status, bdevs = _FindDisks(nodes_ip, disks)
+ if not status:
+ return status, bdevs
+
+ min_resync = 100
+ alldone = True
+ failure = False
+ for rd in bdevs:
+ stats = rd.GetProcStatus()
+ if not (stats.is_connected or stats.is_in_resync):
+ failure = True
+ break
+ alldone = alldone and (not stats.is_in_resync)
+ if stats.sync_percent is not None:
+ min_resync = min(min_resync, stats.sync_percent)
+ return (not failure, (alldone, min_resync))
+
+
+def PowercycleNode(hypervisor_type):
+ """Hard-powercycle the node.
+
+ Because we need to return first, and schedule the powercycle in the
+ background, we won't be able to report failures nicely.
+
+ """
+ hyper = hypervisor.GetHypervisor(hypervisor_type)
+ try:
+ pid = os.fork()
+ except OSError, err:
+ # if we can't fork, we'll pretend that we're in the child process
+ pid = 0
+ if pid > 0:
+ return (True, "Reboot scheduled in 5 seconds")
+ time.sleep(5)
+ hyper.PowercycleNode()
+
+
class HooksRunner(object):
"""Hook runner.
#logging.exception("Error while closing fd %s", fd)
pass
- return result == 0, output
+ return result == 0, utils.SafeEncode(output.strip())
def RunHooks(self, hpath, phase, env):
"""Run the scripts in the hooks directory.
node nor not
@type iv_name: str
@param iv_name: the instance-visible name of the
- device, as in L{objects.Disk.iv_name}
+ device, as in objects.Disk.iv_name
@rtype: None
projects / ganeti-local / blobdiff