import errno
import re
import subprocess
+import random
+import logging
+import tempfile
+import zlib
+import base64
-from ganeti import logger
from ganeti import errors
from ganeti import utils
from ganeti import ssh
from ganeti import ssconf
-def _GetSshRunner():
- return ssh.SshRunner()
+def _GetConfig():
+ """Simple wrapper to return a SimpleStore.
+ @rtype: L{ssconf.SimpleStore}
+ @return: a SimpleStore instance
+
+ """
+ return ssconf.SimpleStore()
-def StartMaster():
- """Activate local node as master node.
- There are two needed steps for this:
- - run the master script
- - register the cron script
+def _GetSshRunner(cluster_name):
+ """Simple wrapper to return an SshRunner.
+
+ @type cluster_name: str
+ @param cluster_name: the cluster name, which is needed
+ by the SshRunner constructor
+ @rtype: L{ssh.SshRunner}
+ @return: an SshRunner instance
"""
- result = utils.RunCmd([constants.MASTER_SCRIPT, "-d", "start"])
+ return ssh.SshRunner(cluster_name)
- if result.failed:
- logger.Error("could not activate cluster interface with command %s,"
- " error: '%s'" % (result.cmd, result.output))
- return False
- return True
+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
+ @param path: the directory to clean
+ @type exclude: list
+ @param exclude: list of files to be excluded, defaults
+ to the empty list
+
+ """
+ if not os.path.isdir(path):
+ return
+ 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))
+ if full_name in exclude:
+ continue
+ if os.path.isfile(full_name) and not os.path.islink(full_name):
+ utils.RemoveFile(full_name)
+
+def JobQueuePurge():
+ """Removes job queue files and archived jobs.
-def StopMaster():
+ @rtype: None
+
+ """
+ _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
+ _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
+
+
+def GetMasterInfo():
+ """Returns master information.
+
+ This is an utility function to compute master information, either
+ for consumption here or from the node daemon.
+
+ @rtype: tuple
+ @return: (master_netdev, master_ip, master_name) if we have a good
+ configuration, otherwise (None, None, None)
+
+ """
+ try:
+ cfg = _GetConfig()
+ master_netdev = cfg.GetMasterNetdev()
+ master_ip = cfg.GetMasterIP()
+ master_node = cfg.GetMasterNode()
+ except errors.ConfigurationError, err:
+ logging.exception("Cluster configuration incomplete")
+ return (None, None, None)
+ return (master_netdev, master_ip, master_node)
+
+
+def StartMaster(start_daemons):
+ """Activate local node as master node.
+
+ The function will always try activate the IP address of the master
+ (unless someone else has it). It will also start the master daemons,
+ based on the start_daemons parameter.
+
+ @type start_daemons: boolean
+ @param start_daemons: whther to also start the master
+ daemons (ganeti-masterd and ganeti-rapi)
+ @rtype: None
+
+ """
+ ok = True
+ master_netdev, master_ip, _ = GetMasterInfo()
+ if not master_netdev:
+ return False
+
+ if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
+ if utils.OwnIpAddress(master_ip):
+ # we already have the ip:
+ logging.debug("Already started")
+ else:
+ logging.error("Someone else has the master ip, not activating")
+ ok = False
+ else:
+ result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
+ "dev", master_netdev, "label",
+ "%s:0" % master_netdev])
+ if result.failed:
+ logging.error("Can't activate master IP: %s", result.output)
+ ok = False
+
+ result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
+ "-s", master_ip, master_ip])
+ # we'll ignore the exit code of arping
+
+ # and now start the master and rapi daemons
+ if start_daemons:
+ for daemon in 'ganeti-masterd', 'ganeti-rapi':
+ result = utils.RunCmd([daemon])
+ if result.failed:
+ logging.error("Can't start daemon %s: %s", daemon, result.output)
+ ok = False
+ return ok
+
+
+def StopMaster(stop_daemons):
"""Deactivate this node as master.
- This runs the master stop script.
+ The function will always try to deactivate the IP address of the
+ master. It will also stop the master daemons depending on the
+ stop_daemons parameter.
+
+ @type stop_daemons: boolean
+ @param stop_daemons: whether to also stop the master daemons
+ (ganeti-masterd and ganeti-rapi)
+ @rtype: None
"""
- result = utils.RunCmd([constants.MASTER_SCRIPT, "-d", "stop"])
+ master_netdev, master_ip, _ = GetMasterInfo()
+ if not master_netdev:
+ return False
+ result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
+ "dev", master_netdev])
if result.failed:
- logger.Error("could not deactivate cluster interface with command %s,"
- " error: '%s'" % (result.cmd, result.output))
- return False
+ logging.error("Can't remove the master IP, error: %s", result.output)
+ # but otherwise ignore the failure
+
+ if stop_daemons:
+ # stop/kill the rapi and the master daemon
+ for daemon in constants.RAPI_PID, constants.MASTERD_PID:
+ utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
return True
- adds the ssh private key to the user
- adds the ssh public key to the users' authorized_keys file
+ @type dsa: str
+ @param dsa: the DSA private key to write
+ @type dsapub: str
+ @param dsapub: the DSA public key to write
+ @type rsa: str
+ @param rsa: the RSA private key to write
+ @type rsapub: str
+ @param rsapub: the RSA public key to write
+ @type sshkey: str
+ @param sshkey: the SSH private key to write
+ @type sshpub: str
+ @param sshpub: the SSH public key to write
+ @rtype: boolean
+ @return: the success of the operation
+
"""
sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
(constants.SSH_HOST_RSA_PUB, rsapub, 0644),
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
mkdir=True)
except errors.OpExecError, err:
- logger.Error("Error while processing user ssh files: %s" % err)
+ logging.exception("Error while processing user ssh files")
return False
for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
def LeaveCluster():
- """Cleans up the current node and prepares it to be removed from the cluster.
+ """Cleans up and remove the current node.
+
+ This function cleans up and prepares the current node to be removed
+ from the cluster.
+
+ If processing is successful, then it raises an
+ L{errors.QuitGanetiException} which is used as a special case to
+ shutdown the node daemon.
"""
- if os.path.isdir(constants.DATA_DIR):
- for rel_name in utils.ListVisibleFiles(constants.DATA_DIR):
- full_name = os.path.join(constants.DATA_DIR, rel_name)
- if os.path.isfile(full_name) and not os.path.islink(full_name):
- utils.RemoveFile(full_name)
+ _CleanDirectory(constants.DATA_DIR)
+ JobQueuePurge()
try:
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
- except errors.OpExecError, err:
- logger.Error("Error while processing ssh files: %s" % err)
+ except errors.OpExecError:
+ logging.exception("Error while processing ssh files")
return
f = open(pub_key, 'r')
utils.RemoveFile(priv_key)
utils.RemoveFile(pub_key)
+ # Return a reassuring string to the caller, and quit
+ raise errors.QuitGanetiException(False, 'Shutdown scheduled')
+
-def GetNodeInfo(vgname):
+def GetNodeInfo(vgname, hypervisor_type):
"""Gives back a hash with different informations about the node.
- Returns:
- { 'vg_size' : xxx, 'vg_free' : xxx, 'memory_domain0': xxx,
- 'memory_free' : xxx, 'memory_total' : xxx }
- where
- vg_size is the size of the configured volume group in MiB
- vg_free is the free size of the volume group in MiB
- memory_dom0 is the memory allocated for domain0 in MiB
- memory_free is the currently available (free) ram in MiB
- memory_total is the total number of ram in MiB
+ @type vgname: C{string}
+ @param vgname: 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
+ @rtype: C{dict}
+ @return: dictionary with the following keys:
+ - vg_size is the size of the configured volume group in MiB
+ - vg_free is the free size of the volume group in MiB
+ - memory_dom0 is the memory allocated for domain0 in MiB
+ - memory_free is the currently available (free) ram in MiB
+ - memory_total is the total number of ram in MiB
"""
outputarray = {}
outputarray['vg_size'] = vginfo['vg_size']
outputarray['vg_free'] = vginfo['vg_free']
- hyper = hypervisor.GetHypervisor()
+ hyper = hypervisor.GetHypervisor(hypervisor_type)
hyp_info = hyper.GetNodeInfo()
if hyp_info is not None:
outputarray.update(hyp_info)
return outputarray
-def VerifyNode(what):
+def VerifyNode(what, cluster_name):
"""Verify the status of the local node.
- Args:
- what - a dictionary of things to check:
- 'filelist' : list of files for which to compute checksums
- 'nodelist' : list of nodes we should check communication with
- 'hypervisor': run the hypervisor-specific verify
+ Based on the input L{what} parameter, various checks are done on the
+ local node.
+
+ If the I{filelist} key is present, this list of
+ files is checksummed and the file/checksum pairs are returned.
- Requested files on local node are checksummed and the result returned.
+ If the I{nodelist} key is present, we check that we have
+ connectivity via ssh with the target nodes (and check the hostname
+ report).
- The nodelist is traversed, with the following checks being made
- for each node:
- - known_hosts key correct
- - correct resolving of node name (target node returns its own hostname
- by ssh-execution of 'hostname', result compared against name in list.
+ If the I{node-net-test} key is present, we check that we have
+ connectivity to the given nodes via both primary IP and, if
+ applicable, secondary IPs.
+
+ @type what: C{dict}
+ @param what: a dictionary of things to check:
+ - filelist: list of files for which to compute checksums
+ - nodelist: list of nodes we should check ssh communication with
+ - node-net-test: list of nodes we should check node daemon port
+ connectivity with
+ - hypervisor: list with hypervisors to run the verify for
+ @rtype: dict
+ @return: a dictionary with the same keys as the input dict, and
+ values representing the result of the checks
"""
result = {}
- if 'hypervisor' in what:
- result['hypervisor'] = hypervisor.GetHypervisor().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'] = {}
- for node in what['nodelist']:
- success, message = _GetSshRunner().VerifyNodeHostname(node)
+ 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
+ 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[constants.NV_NODENETTEST]:
+ if name == my_name:
+ my_pip = pip
+ my_sip = sip
+ break
+ if not my_pip:
+ 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[constants.NV_NODENETTEST]:
+ fail = []
+ if not utils.TcpPing(pip, port, source=my_pip):
+ fail.append("primary")
+ if sip != pip:
+ if not utils.TcpPing(sip, port, source=my_sip):
+ fail.append("secondary")
+ if 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
+
+ if constants.NV_HVINFO in what:
+ hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
+ result[constants.NV_HVINFO] = hyper.GetNodeInfo()
+
return result
def GetVolumeList(vg_name):
"""Compute list of logical volumes and their size.
- Returns:
- dictionary of all partions (key) with their size (in MiB), inactive
- and online status:
- {'test1': ('20.06', True, True)}
+ @type vg_name: str
+ @param vg_name: the volume group whose LVs we should list
+ @rtype: dict
+ @return:
+ dictionary of all partions (key) with value being a tuple of
+ their size (in MiB), inactive and online status::
+
+ {'test1': ('20.06', True, True)}
+
+ in case of errors, a string is returned with the error
+ details.
"""
lvs = {}
"--separator=%s" % sep,
"-olv_name,lv_size,lv_attr", vg_name])
if result.failed:
- logger.Error("Failed to list logical volumes, lvs output: %s" %
- result.output)
+ logging.error("Failed to list logical volumes, lvs output: %s",
+ result.output)
return result.output
+ valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
for line in result.stdout.splitlines():
- line = line.strip().rstrip(sep)
- name, size, attr = line.split(sep)
- if len(attr) != 6:
- attr = '------'
+ line = line.strip()
+ match = valid_line_re.match(line)
+ if not match:
+ logging.error("Invalid line returned from lvs output: '%s'", line)
+ continue
+ name, size, attr = match.groups()
inactive = attr[4] == '-'
online = attr[5] == 'o'
lvs[name] = (size, inactive, online)
def ListVolumeGroups():
"""List the volume groups and their size.
- Returns:
- Dictionary with keys volume name and values the size of the volume
+ @rtype: dict
+ @return: dictionary with keys volume name and values the
+ size of the volume
"""
return utils.ListVolumeGroups()
def NodeVolumes():
"""List all volumes on this node.
+ @rtype: list
+ @return:
+ A list of dictionaries, each having four keys:
+ - name: the logical volume name,
+ - size: the size of the logical volume
+ - dev: the physical device on which the LV lives
+ - vg: the volume group to which it belongs
+
+ In case of errors, we return an empty list and log the
+ error.
+
+ Note that since a logical volume can live on multiple physical
+ volumes, the resulting list might include a logical volume
+ multiple times.
+
"""
result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
"--separator=|",
"--options=lv_name,lv_size,devices,vg_name"])
if result.failed:
- logger.Error("Failed to list logical volumes, lvs output: %s" %
- result.output)
- return {}
+ logging.error("Failed to list logical volumes, lvs output: %s",
+ result.output)
+ return []
def parse_dev(dev):
if '(' in dev:
'vg': line[3].strip(),
}
- return [map_line(line.split('|')) for line in result.stdout.splitlines()]
+ return [map_line(line.split('|')) for line in result.stdout.splitlines()
+ if line.count('|') >= 3]
def BridgesExist(bridges_list):
"""Check if a list of bridges exist on the current node.
- Returns:
- True if all of them exist, false otherwise
+ @rtype: boolean
+ @return: C{True} if all of them exist, C{False} otherwise
"""
for bridge in bridges_list:
return True
-def GetInstanceList():
+def GetInstanceList(hypervisor_list):
"""Provides a list of instances.
- Returns:
- A list of all running instances on the current node
+ @type hypervisor_list: list
+ @param hypervisor_list: the list of hypervisors to query information
+
+ @rtype: list
+ @return: a list of all running instances on the current node
- instance1.example.com
- instance2.example.com
"""
- try:
- names = hypervisor.GetHypervisor().ListInstances()
- except errors.HypervisorError, err:
- logger.Error("error enumerating instances: %s" % str(err))
- raise
+ results = []
+ for hname in hypervisor_list:
+ try:
+ names = hypervisor.GetHypervisor(hname).ListInstances()
+ 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 names
+ return results
-def GetInstanceInfo(instance):
+def GetInstanceInfo(instance, hname):
"""Gives back the informations about an instance as a dictionary.
- Args:
- instance: name of the instance (ex. instance1.example.com)
+ @type instance: string
+ @param instance: the instance name
+ @type hname: string
+ @param hname: the hypervisor type of the instance
- Returns:
- { 'memory' : 511, 'state' : '-b---', 'time' : 3188.8, }
- where
- memory: memory size of instance (int)
- state: xen state of instance (string)
- time: cpu time of instance (float)
+ @rtype: dict
+ @return: dictionary with the following keys:
+ - memory: memory size of instance (int)
+ - state: xen state of instance (string)
+ - time: cpu time of instance (float)
"""
output = {}
- iinfo = hypervisor.GetHypervisor().GetInstanceInfo(instance)
+ iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
if iinfo is not None:
output['memory'] = iinfo[2]
output['state'] = iinfo[4]
return output
-def GetAllInstancesInfo():
+def GetAllInstancesInfo(hypervisor_list):
"""Gather data about all instances.
- This is the equivalent of `GetInstanceInfo()`, except that it
+ This is the equivalent of L{GetInstanceInfo}, except that it
computes data for all instances at once, thus being faster if one
needs data about more than one instance.
- Returns: a dictionary of dictionaries, keys being the instance name,
- and with values:
- { 'memory' : 511, 'state' : '-b---', 'time' : 3188.8, }
- where
- memory: memory size of instance (int)
- state: xen state of instance (string)
- time: cpu time of instance (float)
- vcpus: the number of cpus
+ @type hypervisor_list: list
+ @param hypervisor_list: list of hypervisors to query for instance data
+
+ @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)
+ - time: cpu time of instance (float)
+ - vcpus: the number of vcpus
"""
output = {}
- iinfo = hypervisor.GetHypervisor().GetAllInstancesInfo()
- if iinfo:
- for name, inst_id, memory, vcpus, state, times in iinfo:
- output[name] = {
- 'memory': memory,
- 'vcpus': vcpus,
- 'state': state,
- 'time': times,
- }
+ for hname in hypervisor_list:
+ iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
+ if iinfo:
+ for name, inst_id, memory, vcpus, state, times in iinfo:
+ value = {
+ 'memory': memory,
+ 'vcpus': vcpus,
+ 'state': state,
+ 'time': times,
+ }
+ if name in output and output[name] != value:
+ raise errors.HypervisorError("Instance %s running duplicate"
+ " with different parameters" % name)
+ output[name] = value
return output
-def AddOSToInstance(instance, os_disk, swap_disk):
+def AddOSToInstance(instance):
"""Add an OS to an instance.
- Args:
- instance: the instance object
- os_disk: the instance-visible name of the os device
- swap_disk: the instance-visible name of the swap device
+ @type instance: L{objects.Instance}
+ @param instance: Instance whose OS is to be installed
+ @rtype: boolean
+ @return: the success of the operation
"""
inst_os = OSFromDisk(instance.os)
- create_script = inst_os.create_script
-
- os_device = instance.FindDisk(os_disk)
- if os_device is None:
- logger.Error("Can't find this device-visible name '%s'" % os_disk)
- return False
-
- swap_device = instance.FindDisk(swap_disk)
- if swap_device is None:
- logger.Error("Can't find this device-visible name '%s'" % swap_disk)
- return False
-
- real_os_dev = _RecursiveFindBD(os_device)
- if real_os_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(os_device))
- real_os_dev.Open()
-
- real_swap_dev = _RecursiveFindBD(swap_device)
- if real_swap_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(swap_device))
- real_swap_dev.Open()
+ create_env = OSEnvironment(instance)
logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
instance.name, int(time.time()))
- if not os.path.exists(constants.LOG_OS_DIR):
- os.mkdir(constants.LOG_OS_DIR, 0750)
-
- command = utils.BuildShellCmd("cd %s && %s -i %s -b %s -s %s &>%s",
- inst_os.path, create_script, instance.name,
- real_os_dev.dev_path, real_swap_dev.dev_path,
- logfile)
- result = utils.RunCmd(command)
+ result = utils.RunCmd([inst_os.create_script], env=create_env,
+ cwd=inst_os.path, output=logfile,)
if result.failed:
- logger.Error("os create command '%s' returned error: %s, logfile: %s,"
- " output: %s" %
- (command, result.fail_reason, logfile, result.output))
+ logging.error("os create command '%s' returned error: %s, logfile: %s,"
+ " output: %s", result.cmd, result.fail_reason, logfile,
+ result.output)
return False
return True
-def RunRenameInstance(instance, old_name, os_disk, swap_disk):
+def RunRenameInstance(instance, old_name):
"""Run the OS rename script for an instance.
- Args:
- instance: the instance object
- old_name: the old name of the instance
- os_disk: the instance-visible name of the os device
- swap_disk: the instance-visible name of the swap device
+ @type instance: L{objects.Instance}
+ @param instance: Instance whose OS is to be installed
+ @type old_name: string
+ @param old_name: previous instance name
+ @rtype: boolean
+ @return: the success of the operation
"""
inst_os = OSFromDisk(instance.os)
- script = inst_os.rename_script
-
- os_device = instance.FindDisk(os_disk)
- if os_device is None:
- logger.Error("Can't find this device-visible name '%s'" % os_disk)
- return False
-
- swap_device = instance.FindDisk(swap_disk)
- if swap_device is None:
- logger.Error("Can't find this device-visible name '%s'" % swap_disk)
- return False
-
- real_os_dev = _RecursiveFindBD(os_device)
- if real_os_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(os_device))
- real_os_dev.Open()
-
- real_swap_dev = _RecursiveFindBD(swap_device)
- if real_swap_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(swap_device))
- real_swap_dev.Open()
+ rename_env = OSEnvironment(instance)
+ rename_env['OLD_INSTANCE_NAME'] = old_name
logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
old_name,
instance.name, int(time.time()))
- if not os.path.exists(constants.LOG_OS_DIR):
- os.mkdir(constants.LOG_OS_DIR, 0750)
-
- command = utils.BuildShellCmd("cd %s && %s -o %s -n %s -b %s -s %s &>%s",
- inst_os.path, script, old_name, instance.name,
- real_os_dev.dev_path, real_swap_dev.dev_path,
- logfile)
- result = utils.RunCmd(command)
+ result = utils.RunCmd([inst_os.rename_script], env=rename_env,
+ cwd=inst_os.path, output=logfile)
if result.failed:
- logger.Error("os create command '%s' returned error: %s"
- " output: %s" %
- (command, result.fail_reason, result.output))
+ logging.error("os create command '%s' returned error: %s output: %s",
+ result.cmd, result.fail_reason, result.output)
return False
return True
def _GetVGInfo(vg_name):
"""Get informations about the volume group.
- Args:
- vg_name: the volume group
-
- Returns:
- { 'vg_size' : xxx, 'vg_free' : xxx, 'pv_count' : xxx }
- where
- vg_size is the total size of the volume group in MiB
- vg_free is the free size of the volume group in MiB
- pv_count are the number of physical disks in that vg
+ @type vg_name: str
+ @param vg_name: the volume group which we query
+ @rtype: dict
+ @return:
+ A dictionary with the following keys:
+ - C{vg_size} is the total size of the volume group in MiB
+ - C{vg_free} is the free size of the volume group in MiB
+ - C{pv_count} are the number of physical disks in that VG
- If an error occurs during gathering of data, we return the same dict
- with keys all set to None.
+ If an error occurs during gathering of data, we return the same dict
+ with keys all set to None.
"""
retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
"--nosuffix", "--units=m", "--separator=:", vg_name])
if retval.failed:
- errmsg = "volume group %s not present" % vg_name
- logger.Error(errmsg)
+ logging.error("volume group %s not present", vg_name)
return retdic
valarr = retval.stdout.strip().rstrip(':').split(':')
if len(valarr) == 3:
"pv_count": int(valarr[2]),
}
except ValueError, err:
- logger.Error("Fail to parse vgs output: %s" % str(err))
+ logging.exception("Fail to parse vgs output")
else:
- logger.Error("vgs output has the wrong number of fields (expected"
- " three): %s" % str(valarr))
+ logging.error("vgs output has the wrong number of fields (expected"
+ " three): %s", str(valarr))
return retdic
-def _GatherBlockDevs(instance):
+def _SymlinkBlockDev(instance_name, device_path, device_name):
+ """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).
+
+ Args:
+ instance_name: the name of the target instance
+ device_path: path of the physical block device, on the node
+ device_name: 'virtual' name of the device
+
+ Returns:
+ absolute path to the disk's symlink
+
+ """
+ link_basename = "%s-%s" % (instance_name, device_name)
+ link_name = os.path.join(constants.DISK_LINKS_DIR, link_basename)
+ try:
+ os.symlink(device_path, link_name)
+ except OSError, e:
+ if e.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 _GatherAndLinkBlockDevs(instance):
"""Set up an instance's block device(s).
This is run on the primary node at instance startup. The block
devices must be already assembled.
+ @type instance: L{objects.Instance}
+ @param instance: the instance whose disks we shoul assemble
+ @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,
+ "disk%d" % 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):
"""Start an instance.
- Args:
- instance - name of instance to start.
+ @type instance: L{objects.Instance}
+ @param instance: the instance object
+ @rtype: boolean
+ @return: whether the startup was successful or not
"""
- running_instances = GetInstanceList()
+ running_instances = GetInstanceList([instance.hypervisor])
if instance.name in running_instances:
return True
- block_devices = _GatherBlockDevs(instance)
- hyper = hypervisor.GetHypervisor()
+ block_devices = _GatherAndLinkBlockDevs(instance)
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
try:
hyper.StartInstance(instance, block_devices, extra_args)
except errors.HypervisorError, err:
- logger.Error("Failed to start instance: %s" % err)
+ logging.exception("Failed to start instance")
return False
return True
def ShutdownInstance(instance):
"""Shut an instance down.
- Args:
- instance - name of instance to shutdown.
+ @note: this functions uses polling with a hardcoded timeout.
+
+ @type instance: L{objects.Instance}
+ @param instance: the instance object
+ @rtype: boolean
+ @return: whether the startup was successful or not
"""
- running_instances = GetInstanceList()
+ hv_name = instance.hypervisor
+ running_instances = GetInstanceList([hv_name])
if instance.name not in running_instances:
return True
- hyper = hypervisor.GetHypervisor()
+ hyper = hypervisor.GetHypervisor(hv_name)
try:
hyper.StopInstance(instance)
except errors.HypervisorError, err:
- logger.Error("Failed to stop instance: %s" % err)
+ logging.error("Failed to stop instance")
return False
# test every 10secs for 2min
- shutdown_ok = False
time.sleep(1)
for dummy in range(11):
- if instance.name not in GetInstanceList():
+ if instance.name not in GetInstanceList([hv_name]):
break
time.sleep(10)
else:
# the shutdown did not succeed
- logger.Error("shutdown of '%s' unsuccessful, using destroy" % instance)
+ logging.error("shutdown of '%s' unsuccessful, using destroy", instance)
try:
hyper.StopInstance(instance, force=True)
except errors.HypervisorError, err:
- logger.Error("Failed to stop instance: %s" % err)
+ logging.exception("Failed to stop instance")
return False
time.sleep(1)
- if instance.name in GetInstanceList():
- logger.Error("could not shutdown instance '%s' even by destroy")
+ if instance.name in GetInstanceList([hv_name]):
+ logging.error("could not shutdown instance '%s' even by destroy",
+ instance.name)
return False
return True
def RebootInstance(instance, reboot_type, extra_args):
"""Reboot an instance.
- Args:
- instance - name of instance to reboot
- reboot_type - how to reboot [soft,hard,full]
+ @type instance: L{objects.Instance}
+ @param instance: the instance object to reboot
+ @type reboot_type: str
+ @param reboot_type: the type of reboot, one the following
+ constants:
+ - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
+ instance OS, do not recreate the VM
+ - L{constants.INSTANCE_REBOOT_HARD}: tear down and
+ restart the VM (at the hypervisor level)
+ - the other reboot type (L{constants.INSTANCE_REBOOT_HARD})
+ is not accepted here, since that mode is handled
+ differently
+ @rtype: boolean
+ @return: the success of the operation
"""
- running_instances = GetInstanceList()
+ running_instances = GetInstanceList([instance.hypervisor])
if instance.name not in running_instances:
- logger.Error("Cannot reboot instance that is not running")
+ logging.error("Cannot reboot instance that is not running")
return False
- hyper = hypervisor.GetHypervisor()
+ hyper = hypervisor.GetHypervisor(instance.hypervisor)
if reboot_type == constants.INSTANCE_REBOOT_SOFT:
try:
hyper.RebootInstance(instance)
except errors.HypervisorError, err:
- logger.Error("Failed to soft reboot instance: %s" % err)
+ logging.exception("Failed to soft reboot instance")
return False
elif reboot_type == constants.INSTANCE_REBOOT_HARD:
try:
ShutdownInstance(instance)
StartInstance(instance, extra_args)
except errors.HypervisorError, err:
- logger.Error("Failed to hard reboot instance: %s" % err)
+ logging.exception("Failed to hard reboot instance")
return False
else:
raise errors.ParameterError("reboot_type invalid")
-
return True
+def MigrateInstance(instance, target, live):
+ """Migrates an instance to another node.
+
+ @type instance: L{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)
+ @rtype: tuple
+ @return: a tuple of (success, msg) where:
+ - succes is a boolean denoting the success/failure of the operation
+ - msg is a string with details in case of failure
+
+ """
+ hyper = hypervisor.GetHypervisor(instance.hypervisor_name)
+
+ 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)
+ return (True, "Migration successfull")
+
+
def CreateBlockDevice(disk, size, owner, on_primary, info):
"""Creates a block device for an instance.
- Args:
- disk: a ganeti.objects.Disk object
- size: the size of the physical underlying device
- owner: a string with the name of the instance
- on_primary: a boolean indicating if it is the primary node or not
- info: string that will be sent to the physical device creation
-
- Returns:
- the new unique_id of the device (this can sometime be
- computed only after creation), or None. On secondary nodes,
- it's not required to return anything.
+ @type disk: L{objects.Disk}
+ @param disk: the object describing the disk we should create
+ @type size: int
+ @param size: the size of the physical underlying device, in MiB
+ @type owner: str
+ @param owner: the name of the instance for which disk is created,
+ used for device cache data
+ @type on_primary: boolean
+ @param on_primary: indicates if it is the primary node or not
+ @type info: string
+ @param info: string that will be sent to the physical device
+ creation, used for example to set (LVM) tags on LVs
+
+ @return: the new unique_id of the device (this can sometime be
+ computed only after creation), or None. On secondary nodes,
+ it's not required to return anything.
"""
clist = []
try:
device = bdev.FindDevice(disk.dev_type, disk.physical_id, clist)
if device is not None:
- logger.Info("removing existing device %s" % disk)
+ logging.info("removing existing device %s", disk)
device.Remove()
except errors.BlockDeviceError, err:
pass
if on_primary or disk.AssembleOnSecondary():
if not device.Assemble():
errorstring = "Can't assemble device after creation"
- logger.Error(errorstring)
+ logging.error(errorstring)
raise errors.BlockDeviceError("%s, very unusual event - check the node"
" daemon logs" % errorstring)
device.SetSyncSpeed(constants.SYNC_SPEED)
def RemoveBlockDevice(disk):
"""Remove a block device.
- This is intended to be called recursively.
+ @note: This is intended to be called recursively.
+
+ @type disk: L{objects.Disk}
+ @param disk: the disk object we should remove
+ @rtype: boolean
+ @return: the success of the operation
"""
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
- logger.Info("Can't attach to device %s in remove" % disk)
+ logging.info("Can't attach to device %s in remove", disk)
rdev = None
if rdev is not None:
r_path = rdev.dev_path
This is run on the primary and secondary nodes for an instance.
- This function is called recursively.
+ @note: this function is called recursively.
- Args:
- disk: a objects.Disk object
- as_primary: if we should make the block device read/write
-
- Returns:
- the assembled device or None (in case no device was assembled)
+ @type disk: L{objects.Disk}
+ @param disk: the disk we try to assemble
+ @type owner: str
+ @param owner: the name of the instance which owns the disk
+ @type as_primary: boolean
+ @param as_primary: if we should make the block device
+ read/write
- If the assembly is not successful, an exception is raised.
+ @return: the assembled device or None (in case no device
+ was assembled)
+ @raise errors.BlockDeviceError: in case there is an error
+ during the activation of the children or the device
+ itself
"""
children = []
if children.count(None) >= mcn:
raise
cdev = None
- logger.Debug("Error in child activation: %s" % str(err))
+ logging.debug("Error in child activation: %s", str(err))
children.append(cdev)
if as_primary or disk.AssembleOnSecondary():
This is a wrapper over _RecursiveAssembleBD.
- Returns:
- a /dev path for primary nodes
- True for secondary nodes
+ @rtype: str or boolean
+ @return: a C{/dev/...} path for primary nodes, and
+ C{True} for secondary nodes
"""
result = _RecursiveAssembleBD(disk, owner, as_primary)
def ShutdownBlockDevice(disk):
"""Shut down a block device.
- First, if the device is assembled (can `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
devices doesn't require that the upper device was active.
+ @type disk: L{objects.Disk}
+ @param disk: the description of the disk we should
+ shutdown
+ @rtype: boolean
+ @return: the success of the operation
+
"""
r_dev = _RecursiveFindBD(disk)
if r_dev is not None:
def MirrorAddChildren(parent_cdev, new_cdevs):
"""Extend a mirrored block device.
+ @type parent_cdev: L{objects.Disk}
+ @param parent_cdev: the disk to which we should add children
+ @type new_cdevs: list of L{objects.Disk}
+ @param new_cdevs: the list of children which we should add
+ @rtype: boolean
+ @return: the success of the operation
+
"""
- parent_bdev = _RecursiveFindBD(parent_cdev, allow_partial=True)
+ parent_bdev = _RecursiveFindBD(parent_cdev)
if parent_bdev is None:
- logger.Error("Can't find parent device")
+ logging.error("Can't find parent device")
return False
new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
if new_bdevs.count(None) > 0:
- logger.Error("Can't find new device(s) to add: %s:%s" %
- (new_bdevs, new_cdevs))
+ logging.error("Can't find new device(s) to add: %s:%s",
+ new_bdevs, new_cdevs)
return False
parent_bdev.AddChildren(new_bdevs)
return True
def MirrorRemoveChildren(parent_cdev, new_cdevs):
"""Shrink a mirrored block device.
+ @type parent_cdev: L{objects.Disk}
+ @param parent_cdev: the disk from which we should remove children
+ @type new_cdevs: list of L{objects.Disk}
+ @param new_cdevs: the list of children which we should remove
+ @rtype: boolean
+ @return: the success of the operation
+
"""
parent_bdev = _RecursiveFindBD(parent_cdev)
if parent_bdev is None:
- logger.Error("Can't find parent in remove children: %s" % parent_cdev)
+ logging.error("Can't find parent in remove children: %s", parent_cdev)
return False
devs = []
for disk in new_cdevs:
if rpath is None:
bd = _RecursiveFindBD(disk)
if bd is None:
- logger.Error("Can't find dynamic device %s while removing children" %
- disk)
+ logging.error("Can't find dynamic device %s while removing children",
+ disk)
return False
else:
devs.append(bd.dev_path)
def GetMirrorStatus(disks):
"""Get the mirroring status of a list of devices.
- Args:
- disks: list of `objects.Disk`
-
- Returns:
- list of (mirror_done, estimated_time) tuples, which
- are the result of bdev.BlockDevice.CombinedSyncStatus()
+ @type disks: list of L{objects.Disk}
+ @param disks: the list of disks which we should query
+ @rtype: disk
+ @return:
+ a list of (mirror_done, estimated_time) tuples, which
+ are the result of L{bdev.BlockDev.CombinedSyncStatus}
+ @raise errors.BlockDeviceError: if any of the disks cannot be
+ found
"""
stats = []
return stats
-def _RecursiveFindBD(disk, allow_partial=False):
+def _RecursiveFindBD(disk):
"""Check if a device is activated.
If so, return informations about the real device.
- Args:
- disk: the objects.Disk instance
- allow_partial: don't abort the find if a child of the
- device can't be found; this is intended to be
- used when repairing mirrors
+ @type disk: L{objects.Disk}
+ @param disk: the disk object we need to find
- Returns:
- None if the device can't be found
- otherwise the device instance
+ @return: None if the device can't be found,
+ otherwise the device instance
"""
children = []
def FindBlockDevice(disk):
"""Check if a device is activated.
- If so, return informations about the real device.
+ If it is, return informations about the real device.
- Args:
- disk: the objects.Disk instance
- Returns:
- None if the device can't be found
- (device_path, major, minor, sync_percent, estimated_time, is_degraded)
+ @type disk: L{objects.Disk}
+ @param disk: the disk to find
+ @rtype: None or tuple
+ @return: None if the disk cannot be found, otherwise a
+ tuple (device_path, major, minor, sync_percent,
+ estimated_time, is_degraded)
"""
rbd = _RecursiveFindBD(disk)
This allows the master to overwrite(!) a file. It will only perform
the operation if the file belongs to a list of configuration files.
+ @type file_name: str
+ @param file_name: the target file name
+ @type data: str
+ @param data: the new contents of the file
+ @type mode: int
+ @param mode: the mode to give the file (can be None)
+ @type uid: int
+ @param uid: the owner of the file (can be -1 for default)
+ @type gid: int
+ @param gid: the group of the file (can be -1 for default)
+ @type atime: float
+ @param atime: the atime to set on the file (can be None)
+ @type mtime: float
+ @param mtime: the mtime to set on the file (can be None)
+ @rtype: boolean
+ @return: the success of the operation; errors are logged
+ in the node daemon log
+
"""
if not os.path.isabs(file_name):
- logger.Error("Filename passed to UploadFile is not absolute: '%s'" %
- file_name)
+ logging.error("Filename passed to UploadFile is not absolute: '%s'",
+ file_name)
return False
allowed_files = [
constants.CLUSTER_CONF_FILE,
constants.ETC_HOSTS,
constants.SSH_KNOWN_HOSTS_FILE,
+ constants.VNC_PASSWORD_FILE,
]
- allowed_files.extend(ssconf.SimpleStore().GetFileList())
+
if file_name not in allowed_files:
- logger.Error("Filename passed to UploadFile not in allowed"
- " upload targets: '%s'" % file_name)
+ logging.error("Filename passed to UploadFile not in allowed"
+ " upload targets: '%s'", file_name)
return False
- 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
+def WriteSsconfFiles(values):
+ """Update all ssconf files.
+
+ Wrapper around the SimpleStore.WriteFiles.
+
+ """
+ ssconf.SimpleStore().WriteFiles(values)
+
+
def _ErrnoOrStr(err):
"""Format an EnvironmentError exception.
- If the `err` argument has an errno attribute, it will be looked up
- and converted into a textual EXXXX description. Otherwise the string
- representation of the error will be returned.
+ If the L{err} argument has an errno attribute, it will be looked up
+ and converted into a textual C{E...} description. Otherwise the
+ string representation of the error will be returned.
+
+ @type err: L{EnvironmentError}
+ @param err: the exception to format
"""
if hasattr(err, 'errno'):
return detail
-def _OSSearch(name, search_path=None):
- """Search for OSes with the given name in the search_path.
-
- Args:
- name: The name of the OS to look for
- search_path: List of dirs to search (defaults to constants.OS_SEARCH_PATH)
-
- Returns:
- The base_dir the OS resides in
-
- """
- if search_path is None:
- search_path = constants.OS_SEARCH_PATH
-
- for dir_name in search_path:
- t_os_dir = os.path.sep.join([dir_name, name])
- if os.path.isdir(t_os_dir):
- return dir_name
-
- return None
-
-
def _OSOndiskVersion(name, os_dir):
"""Compute and return the API version of a given OS.
- This function will try to read the API version of the os given by
+ This function will try to read the API version of the OS given by
the 'name' parameter and residing in the 'os_dir' directory.
- Return value will be either an integer denoting the version or None in the
- case when this is not a valid OS name.
+ @type name: str
+ @param name: the OS name we should look for
+ @type os_dir: str
+ @param os_dir: the directory inwhich we should look for the OS
+ @rtype: int or None
+ @return:
+ Either an integer denoting the version or None in the
+ case when this is not a valid OS name.
+ @raise errors.InvalidOS: if the OS cannot be found
"""
api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
try:
f = open(api_file)
try:
- api_version = f.read(256)
+ api_versions = f.readlines()
finally:
f.close()
except EnvironmentError, err:
raise errors.InvalidOS(name, os_dir, "error while reading the"
" API version (%s)" % _ErrnoOrStr(err))
- api_version = api_version.strip()
+ api_versions = [version.strip() for version in api_versions]
try:
- api_version = int(api_version)
+ api_versions = [int(version) for version in api_versions]
except (TypeError, ValueError), err:
raise errors.InvalidOS(name, os_dir,
"API version is not integer (%s)" % str(err))
- return api_version
+ return api_versions
def DiagnoseOS(top_dirs=None):
"""Compute the validity for all OSes.
- Returns an OS object for each name in all the given top directories
- (if not given defaults to constants.OS_SEARCH_PATH)
-
- Returns:
- list of OS objects
+ @type top_dirs: list
+ @param top_dirs: the list of directories in which to
+ search (if not given defaults to
+ L{constants.OS_SEARCH_PATH})
+ @rtype: list of L{objects.OS}
+ @return: an OS object for each name in all the given
+ directories
"""
if top_dirs is None:
try:
f_names = utils.ListVisibleFiles(dir_name)
except EnvironmentError, err:
- logger.Error("Can't list the OS directory %s: %s" %
- (dir_name, str(err)))
+ logging.exception("Can't list the OS directory %s", dir_name)
break
for name in f_names:
try:
This function will return an OS instance if the given name is a
valid OS name. Otherwise, it will raise an appropriate
- `errors.InvalidOS` exception, detailing why this is not a valid
- OS.
+ L{errors.InvalidOS} exception, detailing why this is not a valid OS.
- Args:
- os_dir: Directory containing the OS scripts. Defaults to a search
- in all the OS_SEARCH_PATH directories.
+ @type base_dir: string
+ @keyword base_dir: Base directory containing OS installations.
+ Defaults to a search in all the OS_SEARCH_PATH dirs.
+ @rtype: L{objects.OS}
+ @return: the OS instance if we find a valid one
+ @raise errors.InvalidOS: if we don't find a valid OS
"""
-
if base_dir is None:
- base_dir = _OSSearch(name)
-
- if base_dir is None:
- raise errors.InvalidOS(name, None, "OS dir not found in search path")
+ os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
+ if os_dir is None:
+ raise errors.InvalidOS(name, None, "OS dir not found in search path")
+ else:
+ os_dir = os.path.sep.join([base_dir, name])
- os_dir = os.path.sep.join([base_dir, name])
- api_version = _OSOndiskVersion(name, os_dir)
+ api_versions = _OSOndiskVersion(name, os_dir)
- if api_version != constants.OS_API_VERSION:
+ if constants.OS_API_VERSION not in api_versions:
raise errors.InvalidOS(name, os_dir, "API version mismatch"
" (found %s want %s)"
- % (api_version, constants.OS_API_VERSION))
+ % (api_versions, constants.OS_API_VERSION))
# OS Scripts dictionary, we will populate it with the actual script names
- os_scripts = {'create': '', 'export': '', 'import': '', 'rename': ''}
+ os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
for script in os_scripts:
os_scripts[script] = os.path.sep.join([os_dir, script])
return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
- create_script=os_scripts['create'],
- export_script=os_scripts['export'],
- import_script=os_scripts['import'],
- rename_script=os_scripts['rename'],
- api_version=api_version)
+ create_script=os_scripts[constants.OS_SCRIPT_CREATE],
+ export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
+ import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
+ rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
+ api_versions=api_versions)
+
+def OSEnvironment(instance, debug=0):
+ """Calculate the environment for an os script.
+
+ @type instance: L{objects.Instance}
+ @param instance: target instance for the os script run
+ @type debug: integer
+ @param debug: debug level (0 or 1, for OS Api 10)
+ @rtype: dict
+ @return: dict of environment variables
+ @raise errors.BlockDeviceError: if the block device
+ cannot be found
+
+ """
+ result = {}
+ result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
+ result['INSTANCE_NAME'] = instance.name
+ result['HYPERVISOR'] = instance.hypervisor
+ result['DISK_COUNT'] = '%d' % len(instance.disks)
+ result['NIC_COUNT'] = '%d' % len(instance.nics)
+ result['DEBUG_LEVEL'] = '%d' % debug
+ for idx, disk in enumerate(instance.disks):
+ real_disk = _RecursiveFindBD(disk)
+ if real_disk is None:
+ raise errors.BlockDeviceError("Block device '%s' is not set up" %
+ 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'
+ 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:
+ result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
+ elif disk.dev_type == constants.LD_FILE:
+ result['DISK_%d_BACKEND_TYPE' % idx] = \
+ 'file:%s' % disk.physical_id[0]
+ for idx, nic in enumerate(instance.nics):
+ result['NIC_%d_MAC' % idx] = nic.mac
+ if nic.ip:
+ result['NIC_%d_IP' % idx] = nic.ip
+ result['NIC_%d_BRIDGE' % idx] = nic.bridge
+ 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):
+ """Grow a stack of block devices.
+
+ This function is called recursively, with the childrens being the
+ first ones to resize.
+
+ @type disk: L{objects.Disk}
+ @param disk: the disk to be grown
+ @rtype: (status, result)
+ @return: a tuple with the status of the operation
+ (True/False), and the errors message if status
+ is False
+
+ """
+ r_dev = _RecursiveFindBD(disk)
+ if r_dev is None:
+ return False, "Cannot find block device %s" % (disk,)
+
+ try:
+ r_dev.Grow(amount)
+ except errors.BlockDeviceError, err:
+ return False, str(err)
+
+ return True, None
def SnapshotBlockDevice(disk):
This function is called recursively, and the snapshot is actually created
just for the leaf lvm backend device.
- Args:
- disk: the disk to be snapshotted
-
- Returns:
- a config entry for the actual lvm device snapshotted.
+ @type disk: L{objects.Disk}
+ @param disk: the disk to be snapshotted
+ @rtype: string
+ @return: snapshot disk path
"""
if disk.children:
(disk.unique_id, disk.dev_type))
-def ExportSnapshot(disk, dest_node, instance):
+def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
"""Export a block device snapshot to a remote node.
- Args:
- disk: the snapshot block device
- dest_node: the node to send the image to
- instance: instance being exported
-
- Returns:
- True if successful, False otherwise.
+ @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 instance: L{objects.Instance}
+ @param instance: the instance object to whom the disk belongs
+ @type cluster_name: str
+ @param cluster_name: the cluster name, needed for SSH hostalias
+ @type idx: int
+ @param idx: the index of the disk in the instance's disk list,
+ used to export to the OS scripts environment
+ @rtype: boolean
+ @return: the success of the operation
"""
+ export_env = OSEnvironment(instance)
+
inst_os = OSFromDisk(instance.os)
export_script = inst_os.export_script
instance.name, int(time.time()))
if not os.path.exists(constants.LOG_OS_DIR):
os.mkdir(constants.LOG_OS_DIR, 0750)
-
- real_os_dev = _RecursiveFindBD(disk)
- if real_os_dev is None:
+ real_disk = _RecursiveFindBD(disk)
+ if real_disk is None:
raise errors.BlockDeviceError("Block device '%s' is not set up" %
str(disk))
- real_os_dev.Open()
+ real_disk.Open()
+
+ export_env['EXPORT_DEVICE'] = real_disk.dev_path
+ export_env['EXPORT_INDEX'] = str(idx)
destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
destfile = disk.physical_id[1]
# the target command is built out of three individual commands,
# which are joined by pipes; we check each individual command for
# valid parameters
-
- expcmd = utils.BuildShellCmd("cd %s; %s -i %s -b %s 2>%s", inst_os.path,
- export_script, instance.name,
- real_os_dev.dev_path, logfile)
+ expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
+ export_script, logfile)
comprcmd = "gzip"
destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
destdir, destdir, destfile)
- remotecmd = _GetSshRunner().BuildCmd(dest_node, constants.GANETI_RUNAS,
- destcmd)
+ remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
+ constants.GANETI_RUNAS,
+ destcmd)
# all commands have been checked, so we're safe to combine them
command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
- result = utils.RunCmd(command)
+ result = utils.RunCmd(command, env=export_env)
if result.failed:
- logger.Error("os snapshot export command '%s' returned error: %s"
- " output: %s" %
- (command, result.fail_reason, result.output))
+ logging.error("os snapshot export command '%s' returned error: %s"
+ " output: %s", command, result.fail_reason, result.output)
return False
return True
def FinalizeExport(instance, snap_disks):
"""Write out the export configuration information.
- Args:
- instance: instance configuration
- snap_disks: snapshot block devices
+ @type instance: L{objects.Instance}
+ @param instance: the instance which we export, used for
+ saving configuration
+ @type snap_disks: list of L{objects.Disk}
+ @param snap_disks: list of snapshot block devices, which
+ will be used to get the actual name of the dump file
- Returns:
- False in case of error, True otherwise.
+ @rtype: boolean
+ @return: the success of the operation
"""
destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
config.add_section(constants.INISECT_INS)
config.set(constants.INISECT_INS, 'name', instance.name)
- config.set(constants.INISECT_INS, 'memory', '%d' % instance.memory)
- config.set(constants.INISECT_INS, 'vcpus', '%d' % instance.vcpus)
+ config.set(constants.INISECT_INS, 'memory', '%d' %
+ instance.beparams[constants.BE_MEMORY])
+ config.set(constants.INISECT_INS, 'vcpus', '%d' %
+ instance.beparams[constants.BE_VCPUS])
config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
+
+ 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)
+ 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_total = 0
for disk_count, disk in enumerate(snap_disks):
- 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()
-
+ 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_total)
+
+ utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
+ data=config.Dumps())
shutil.rmtree(finaldestdir, True)
shutil.move(destdir, finaldestdir)
def ExportInfo(dest):
"""Get export configuration information.
- Args:
- dest: directory containing the export
+ @type dest: str
+ @param dest: directory containing the export
- Returns:
- A serializable config file containing the export info.
+ @rtype: L{objects.SerializableConfigParser}
+ @return: a serializable config file containing the
+ export info
"""
cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
return config
-def ImportOSIntoInstance(instance, os_disk, swap_disk, src_node, src_image):
+def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
"""Import an os image into an instance.
- Args:
- instance: the instance object
- os_disk: the instance-visible name of the os device
- swap_disk: the instance-visible name of the swap device
- src_node: node holding the source image
- src_image: path to the source image on src_node
-
- Returns:
- False in case of error, True otherwise.
+ @type instance: L{objects.Instance}
+ @param instance: instance to import the disks into
+ @type src_node: string
+ @param src_node: source node for the disk images
+ @type src_images: list of string
+ @param src_images: absolute paths of the disk images
+ @rtype: list of boolean
+ @return: each boolean represent the success of importing the n-th disk
"""
+ import_env = OSEnvironment(instance)
inst_os = OSFromDisk(instance.os)
import_script = inst_os.import_script
- os_device = instance.FindDisk(os_disk)
- if os_device is None:
- logger.Error("Can't find this device-visible name '%s'" % os_disk)
- return False
-
- swap_device = instance.FindDisk(swap_disk)
- if swap_device is None:
- logger.Error("Can't find this device-visible name '%s'" % swap_disk)
- return False
-
- real_os_dev = _RecursiveFindBD(os_device)
- if real_os_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(os_device))
- real_os_dev.Open()
-
- real_swap_dev = _RecursiveFindBD(swap_device)
- if real_swap_dev is None:
- raise errors.BlockDeviceError("Block device '%s' is not set up" %
- str(swap_device))
- real_swap_dev.Open()
-
logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
instance.name, int(time.time()))
if not os.path.exists(constants.LOG_OS_DIR):
os.mkdir(constants.LOG_OS_DIR, 0750)
- destcmd = utils.BuildShellCmd('cat %s', src_image)
- remotecmd = _GetSshRunner().BuildCmd(src_node, constants.GANETI_RUNAS,
- destcmd)
-
comprcmd = "gunzip"
- impcmd = utils.BuildShellCmd("(cd %s; %s -i %s -b %s -s %s &>%s)",
- inst_os.path, import_script, instance.name,
- real_os_dev.dev_path, real_swap_dev.dev_path,
- logfile)
-
- command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
-
- result = utils.RunCmd(command)
-
- if result.failed:
- logger.Error("os import command '%s' returned error: %s"
- " output: %s" %
- (command, result.fail_reason, result.output))
- return False
+ impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
+ import_script, logfile)
+
+ final_result = []
+ for idx, image in enumerate(src_images):
+ if image:
+ destcmd = utils.BuildShellCmd('cat %s', image)
+ remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
+ constants.GANETI_RUNAS,
+ destcmd)
+ command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
+ import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
+ 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)
+ final_result.append(False)
+ else:
+ final_result.append(True)
+ else:
+ final_result.append(True)
- return True
+ return final_result
def ListExports():
"""Return a list of exports currently available on this machine.
+ @rtype: list
+ @return: list of the exports
+
"""
if os.path.isdir(constants.EXPORT_DIR):
return utils.ListVisibleFiles(constants.EXPORT_DIR)
def RemoveExport(export):
"""Remove an existing export from the node.
- Args:
- export: the name of the export to remove
-
- Returns:
- False in case of error, True otherwise.
+ @type export: str
+ @param export: the name of the export to remove
+ @rtype: boolean
+ @return: the success of the operation
"""
target = os.path.join(constants.EXPORT_DIR, export)
def RenameBlockDevices(devlist):
"""Rename a list of block devices.
- The devlist argument is a list of tuples (disk, new_logical,
- new_physical). The return value will be a combined boolean result
- (True only if all renames succeeded).
+ @type devlist: list of tuples
+ @param devlist: list of tuples of the form (disk,
+ new_logical_id, new_physical_id); disk is an
+ L{objects.Disk} object describing the current disk,
+ and new logical_id/physical_id is the name we
+ rename it to
+ @rtype: boolean
+ @return: True if all renames succeeded, False otherwise
"""
result = True
# cache? for now, we only lose lvm data when we rename, which
# is less critical than DRBD or MD
except errors.BlockDeviceError, err:
- logger.Error("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
default file_storage_dir stored in SimpleStore. Only paths under that
directory are allowed.
- Args:
- file_storage_dir: string with path
-
- Returns:
- normalized file_storage_dir (string) if valid, None otherwise
+ @type file_storage_dir: str
+ @param file_storage_dir: the path to check
+
+ @return: the normalized path if valid, None otherwise
"""
+ cfg = _GetConfig()
file_storage_dir = os.path.normpath(file_storage_dir)
- base_file_storage_dir = ssconf.SimpleStore().GetFileStorageDir()
+ base_file_storage_dir = cfg.GetFileStorageDir()
if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
base_file_storage_dir):
- logger.Error("file storage directory '%s' is not under base file"
- " storage directory '%s'" %
- (file_storage_dir, base_file_storage_dir))
+ logging.error("file storage directory '%s' is not under base file"
+ " storage directory '%s'",
+ file_storage_dir, base_file_storage_dir)
return None
return file_storage_dir
def CreateFileStorageDir(file_storage_dir):
"""Create file storage directory.
- Args:
- file_storage_dir: string containing the path
+ @type file_storage_dir: str
+ @param file_storage_dir: directory to create
- Returns:
- tuple with first element a boolean indicating wheter dir
- creation was successful or not
+ @rtype: tuple
+ @return: tuple with first element a boolean indicating wheter dir
+ creation was successful or not
"""
file_storage_dir = _TransformFileStorageDir(file_storage_dir)
else:
if os.path.exists(file_storage_dir):
if not os.path.isdir(file_storage_dir):
- logger.Error("'%s' is not a directory" % file_storage_dir)
+ logging.error("'%s' is not a directory", file_storage_dir)
result = False,
else:
try:
os.makedirs(file_storage_dir, 0750)
except OSError, err:
- logger.Error("Cannot create file storage directory '%s': %s" %
- (file_storage_dir, err))
+ logging.error("Cannot create file storage directory '%s': %s",
+ file_storage_dir, err)
result = False,
return result
Remove it only if it's empty. If not log an error and return.
- Args:
- file_storage_dir: string containing the path
-
- Returns:
- tuple with first element a boolean indicating wheter dir
- removal was successful or not
+ @type file_storage_dir: str
+ @param file_storage_dir: the directory we should cleanup
+ @rtype: tuple (success,)
+ @return: tuple of one element, C{success}, denoting
+ whether the operation was successfull
"""
file_storage_dir = _TransformFileStorageDir(file_storage_dir)
else:
if os.path.exists(file_storage_dir):
if not os.path.isdir(file_storage_dir):
- logger.Error("'%s' is not a directory" % file_storage_dir)
+ logging.error("'%s' is not a directory", file_storage_dir)
result = False,
# deletes dir only if empty, otherwise we want to return False
try:
os.rmdir(file_storage_dir)
except OSError, err:
- logger.Error("Cannot remove file storage directory '%s': %s" %
- (file_storage_dir, err))
+ logging.exception("Cannot remove file storage directory '%s'",
+ file_storage_dir)
result = False,
return result
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
"""Rename the file storage directory.
- Args:
- old_file_storage_dir: string containing the old path
- new_file_storage_dir: string containing the new path
-
- Returns:
- tuple with first element a boolean indicating wheter dir
- rename was successful or not
+ @type old_file_storage_dir: str
+ @param old_file_storage_dir: the current path
+ @type new_file_storage_dir: str
+ @param new_file_storage_dir: the name we should rename to
+ @rtype: tuple (success,)
+ @return: tuple of one element, C{success}, denoting
+ whether the operation was successful
"""
old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
try:
os.rename(old_file_storage_dir, new_file_storage_dir)
except OSError, err:
- logger.Error("Cannot rename '%s' to '%s': %s"
- % (old_file_storage_dir, new_file_storage_dir, err))
+ logging.exception("Cannot rename '%s' to '%s'",
+ old_file_storage_dir, new_file_storage_dir)
result = False,
else:
- logger.Error("'%s' is not a directory" % old_file_storage_dir)
+ logging.error("'%s' is not a directory", old_file_storage_dir)
result = False,
else:
if os.path.exists(old_file_storage_dir):
- logger.Error("Cannot rename '%s' to '%s'. Both locations exist." %
- old_file_storage_dir, new_file_storage_dir)
+ logging.error("Cannot rename '%s' to '%s'. Both locations exist.",
+ old_file_storage_dir, new_file_storage_dir)
result = False,
return result
+def _IsJobQueueFile(file_name):
+ """Checks whether the given filename is in the queue directory.
+
+ @type file_name: str
+ @param file_name: the file name we should check
+ @rtype: boolean
+ @return: whether the file is under the queue directory
+
+ """
+ queue_dir = os.path.normpath(constants.QUEUE_DIR)
+ result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
+
+ if not result:
+ logging.error("'%s' is not a file in the queue directory",
+ file_name)
+
+ return result
+
+
+def JobQueueUpdate(file_name, content):
+ """Updates a file in the queue directory.
+
+ This is just a wrapper over L{utils.WriteFile}, with proper
+ checking.
+
+ @type file_name: str
+ @param file_name: the job file name
+ @type content: str
+ @param content: the new job contents
+ @rtype: boolean
+ @return: the success of the operation
+
+ """
+ if not _IsJobQueueFile(file_name):
+ return False
+
+ # Write and replace the file atomically
+ utils.WriteFile(file_name, data=_Decompress(content))
+
+ return True
+
+
+def JobQueueRename(old, new):
+ """Renames a job queue file.
+
+ This is just a wrapper over os.rename with proper checking.
+
+ @type old: str
+ @param old: the old (actual) file name
+ @type new: str
+ @param new: the desired file name
+ @rtype: boolean
+ @return: the success of the operation
+
+ """
+ if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)):
+ return False
+
+ utils.RenameFile(old, new, mkdir=True)
+
+ return True
+
+
+def JobQueueSetDrainFlag(drain_flag):
+ """Set the drain flag for the queue.
+
+ This will set or unset the queue drain flag.
+
+ @type drain_flag: boolean
+ @param drain_flag: if True, will set the drain flag, otherwise reset it.
+ @rtype: boolean
+ @return: always True
+ @warning: the function always returns True
+
+ """
+ if drain_flag:
+ utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
+ else:
+ utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
+
+ return True
+
+
+def CloseBlockDevices(disks):
+ """Closes the given block devices.
+
+ This means they will be switched to secondary mode (in case of
+ DRBD).
+
+ @type disks: list of L{objects.Disk}
+ @param disks: the list of disks to be closed
+ @rtype: tuple (success, message)
+ @return: a tuple of success and message, where success
+ indicates the succes of the operation, and message
+ which will contain the error details in case we
+ failed
+
+ """
+ bdevs = []
+ for cf in disks:
+ rd = _RecursiveFindBD(cf)
+ if rd is None:
+ return (False, "Can't find device %s" % cf)
+ bdevs.append(rd)
+
+ msg = []
+ for rd in bdevs:
+ try:
+ rd.Close()
+ except errors.BlockDeviceError, err:
+ msg.append(str(err))
+ if msg:
+ return (False, "Can't make devices secondary: %s" % ",".join(msg))
+ else:
+ return (True, "All devices secondary")
+
+
+def ValidateHVParams(hvname, hvparams):
+ """Validates the given hypervisor parameters.
+
+ @type hvname: string
+ @param hvname: the hypervisor name
+ @type hvparams: dict
+ @param hvparams: the hypervisor parameters to be validated
+ @rtype: tuple (success, message)
+ @return: a tuple of success and message, where success
+ indicates the succes of the operation, and message
+ which will contain the error details in case we
+ failed
+
+ """
+ try:
+ hv_type = hypervisor.GetHypervisor(hvname)
+ hv_type.ValidateParameters(hvparams)
+ return (True, "Validation passed")
+ except errors.HypervisorError, err:
+ 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")
+
+
class HooksRunner(object):
"""Hook runner.
- This class is instantiated on the node side (ganeti-noded) and not on
- the master side.
+ This class is instantiated on the node side (ganeti-noded) and not
+ on the master side.
"""
RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
def __init__(self, hooks_base_dir=None):
"""Constructor for hooks runner.
- Args:
- - hooks_base_dir: if not None, this overrides the
- constants.HOOKS_BASE_DIR (useful for unittests)
- - logs_base_dir: if not None, this overrides the
- constants.LOG_HOOKS_DIR (useful for unittests)
- - logging: enable or disable logging of script output
+ @type hooks_base_dir: str or None
+ @param hooks_base_dir: if not None, this overrides the
+ L{constants.HOOKS_BASE_DIR} (useful for unittests)
"""
if hooks_base_dir is None:
def ExecHook(script, env):
"""Exec one hook script.
- Args:
- - phase: the phase
- - script: the full path to the script
- - env: the environment with which to exec the script
+ @type script: str
+ @param script: the full path to the script
+ @type env: dict
+ @param env: the environment with which to exec the script
+ @rtype: tuple (success, message)
+ @return: a tuple of success and message, where success
+ indicates the succes of the operation, and message
+ which will contain the error details in case we
+ failed
"""
# exec the process using subprocess and log the output
fd.close()
except EnvironmentError, err:
# just log the error
- #logger.Error("While closing fd %s: %s" % (fd, err))
+ #logging.exception("Error while closing fd %s", fd)
pass
return result == 0, output
def RunHooks(self, hpath, phase, env):
"""Run the scripts in the hooks directory.
- This method will not be usually overriden by child opcodes.
+ @type hpath: str
+ @param hpath: the path to the hooks directory which
+ holds the scripts
+ @type phase: str
+ @param phase: either L{constants.HOOKS_PHASE_PRE} or
+ L{constants.HOOKS_PHASE_POST}
+ @type env: dict
+ @param env: dictionary with the environment for the hook
+ @rtype: list
+ @return: list of 3-element tuples:
+ - script path
+ - script result, either L{constants.HKR_SUCCESS} or
+ L{constants.HKR_FAIL}
+ - output of the script
+
+ @raise errors.ProgrammerError: for invalid input
+ parameters
"""
if phase == constants.HOOKS_PHASE_PRE:
try:
dir_contents = utils.ListVisibleFiles(dir_name)
except OSError, err:
- # must log
+ # FIXME: must log output in case of failures
return rr
# we use the standard python sort order,
return rr
+class IAllocatorRunner(object):
+ """IAllocator runner.
+
+ This class is instantiated on the node side (ganeti-noded) and not on
+ the master side.
+
+ """
+ def Run(self, name, idata):
+ """Run an iallocator script.
+
+ @type name: str
+ @param name: the iallocator script name
+ @type idata: str
+ @param idata: the allocator input data
+
+ @rtype: tuple
+ @return: four element tuple of:
+ - run status (one of the IARUN_ constants)
+ - stdout
+ - stderr
+ - fail reason (as from L{utils.RunResult})
+
+ """
+ alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
+ os.path.isfile)
+ if alloc_script is None:
+ return (constants.IARUN_NOTFOUND, None, None, None)
+
+ fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
+ try:
+ os.write(fd, idata)
+ os.close(fd)
+ result = utils.RunCmd([alloc_script, fin_name])
+ if result.failed:
+ return (constants.IARUN_FAILURE, result.stdout, result.stderr,
+ result.fail_reason)
+ finally:
+ os.unlink(fin_name)
+
+ return (constants.IARUN_SUCCESS, result.stdout, result.stderr, None)
+
+
class DevCacheManager(object):
"""Simple class for managing a cache of block device information.
"""Converts a /dev/name path to the cache file name.
This replaces slashes with underscores and strips the /dev
- prefix. It then returns the full path to the cache file
+ prefix. It then returns the full path to the cache file.
+
+ @type dev_path: str
+ @param dev_path: the C{/dev/} path name
+ @rtype: str
+ @return: the converted path name
"""
if dev_path.startswith(cls._DEV_PREFIX):
def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
"""Updates the cache information for a given device.
+ @type dev_path: str
+ @param dev_path: the pathname of the device
+ @type owner: str
+ @param owner: the owner (instance name) of the device
+ @type on_primary: bool
+ @param on_primary: whether this is the primary
+ node nor not
+ @type iv_name: str
+ @param iv_name: the instance-visible name of the
+ device, as in objects.Disk.iv_name
+
+ @rtype: None
+
"""
if dev_path is None:
- logger.Error("DevCacheManager.UpdateCache got a None dev_path")
+ logging.error("DevCacheManager.UpdateCache got a None dev_path")
return
fpath = cls._ConvertPath(dev_path)
if on_primary:
try:
utils.WriteFile(fpath, data=fdata)
except EnvironmentError, err:
- logger.Error("Can't update bdev cache for %s, error %s" %
- (dev_path, str(err)))
+ logging.exception("Can't update bdev cache for %s", dev_path)
@classmethod
def RemoveCache(cls, dev_path):
"""Remove data for a dev_path.
+ This is just a wrapper over L{utils.RemoveFile} with a converted
+ path name and logging.
+
+ @type dev_path: str
+ @param dev_path: the pathname of the device
+
+ @rtype: None
+
"""
if dev_path is None:
- logger.Error("DevCacheManager.RemoveCache got a None dev_path")
+ logging.error("DevCacheManager.RemoveCache got a None dev_path")
return
fpath = cls._ConvertPath(dev_path)
try:
utils.RemoveFile(fpath)
except EnvironmentError, err:
- logger.Error("Can't update bdev cache for %s, error %s" %
- (dev_path, str(err)))
+ logging.exception("Can't update bdev cache for %s", dev_path)
projects / ganeti-local / blobdiff