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
+ msg = "Error while processing user ssh files"
+ logging.exception(msg)
+ return (False, "%s: %s" % (msg, err))
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():
result[constants.NV_VGLIST] = ListVolumeGroups()
if constants.NV_VERSION in what:
- result[constants.NV_VERSION] = constants.PROTOCOL_VERSION
+ 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:
+ logging.warning("Can't get used minors list", exc_info=True)
+ used_minors = []
+ result[constants.NV_DRBDLIST] = used_minors
+
return result
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.
return output
-def AddOSToInstance(instance):
+def InstanceOsAdd(instance):
"""Add an OS to an instance.
@type instance: L{objects.Instance}
@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)
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):
running_instances = GetInstanceList([instance.hypervisor])
if instance.name in running_instances:
- return True
+ return (True, "Already running")
try:
block_devices = _GatherAndLinkBlockDevs(instance)
hyper.StartInstance(instance, block_devices, extra_args)
except errors.BlockDeviceError, err:
logging.exception("Failed to start instance")
- return False
+ return (False, "Block device error: %s" % str(err))
except errors.HypervisorError, err:
logging.exception("Failed to start instance")
_RemoveBlockDevLinks(instance.name, instance.disks)
- return False
+ return (False, "Hypervisor error: %s" % str(err))
- return True
+ return (True, "Instance started successfully")
def ShutdownInstance(instance):
try:
hyper.StopInstance(instance)
except errors.HypervisorError, err:
- logging.error("Failed to stop instance")
+ logging.error("Failed to stop instance: %s" % err)
return False
# test every 10secs for 2min
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")
+ logging.exception("Failed to stop instance: %s" % err)
return False
time.sleep(1)
if instance.name in GetInstanceList([hv_name]):
- logging.error("could not shutdown instance '%s' even by destroy",
+ logging.error("Could not shutdown instance '%s' even by destroy",
instance.name)
return False
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:
+ msg = "Failed to fetch migration information"
+ logging.exception(msg)
+ return (False, '%s: %s' % (msg, err))
+ 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:
+ msg = "Failed to accept instance"
+ logging.exception(msg)
+ return (False, '%s: %s' % (msg, err))
+ 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:
+ msg = "Failed to finalize migration"
+ logging.exception(msg)
+ return (False, '%s: %s' % (msg, err))
+ return (True, "Migration Finalized")
+
+
def MigrateInstance(instance, target, live):
"""Migrates an instance to another node.
- 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)
+ msg = "Failed to migrate instance"
+ logging.exception(msg)
+ return (False, "%s: %s" % (msg, err))
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}
# be assembled
crdev.Open()
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()
- except errors.BlockDeviceError, err:
- pass
+ device = bdev.Create(disk.dev_type, disk.physical_id, clist, size)
+ except errors.GenericError, err:
+ return False, "Can't create block device: %s" % str(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"
+ errorstring = "Can't assemble device after creation, very unusual event"
logging.error(errorstring)
- raise errors.BlockDeviceError("%s, very unusual event - check the node"
- " daemon logs" % errorstring)
+ return False, errorstring
device.SetSyncSpeed(constants.SYNC_SPEED)
if on_primary or disk.OpenOnSecondary():
device.Open(force=True)
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.
@return: the success of the operation
"""
+ msgs = []
+ result = True
try:
rdev = _RecursiveFindBD(disk)
except errors.BlockDeviceError, err:
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: %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 = False
+ result = "no error information"
+ try:
+ result = _RecursiveAssembleBD(disk, owner, as_primary)
+ if isinstance(result, bdev.BlockDev):
+ result = result.dev_path
+ status = True
+ if result == True:
+ status = True
+ except errors.BlockDeviceError, err:
+ result = "Error while assembling disk: %s" % str(err)
+ return (status, result)
-def ShutdownBlockDevice(disk):
+def BlockdevShutdown(disk):
"""Shut down a block device.
First, if the device is assembled (Attach() is successfull), then
@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 True
-def MirrorRemoveChildren(parent_cdev, new_cdevs):
+def BlockdevRemovechildren(parent_cdev, new_cdevs):
"""Shrink a mirrored block device.
@type parent_cdev: L{objects.Disk}
return True
-def GetMirrorStatus(disks):
+def BlockdevGetmirrorstatus(disks):
"""Get the mirroring status of a list of devices.
@type disks: list of L{objects.Disk}
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:
+ return (False, str(err))
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):
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)
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]
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
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:
return True
-def RenameBlockDevices(devlist):
+def BlockdevRename(devlist):
"""Rename a list of block devices.
@type devlist: list of tuples
return True
-def CloseBlockDevices(instance_name, disks):
+def BlockdevClose(instance_name, disks):
"""Closes the given block devices.
This means they will be switched to secondary mode (in case of
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:
+ logging.exception("Failed to go into standalone mode")
+ return (False, "Can't change network configuration: %s" % str(err))
+ 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:
+ return (False, "Can't create symlink: %s" % str(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:
+ return (False, "Can't change network configuration: %s" % str(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:
+ return (False, "Can't change network configuration: %s" % str(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:
+ return (False, "Can't change to primary mode: %s" % str(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))
+
+
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.
projects / ganeti-local / blobdiff