# 02110-1301, USA.
-"""Functions used by the node daemon"""
+"""Functions used by the node daemon
+
+@var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
+ the L{UploadFile} function
+
+"""
+
+# pylint: disable-msg=E1103
+
+# E1103: %s %r has no %r member (but some types could not be
+# inferred), because the _TryOSFromDisk returns either (True, os_obj)
+# or (False, "string") which confuses pylint
import os
from ganeti import ssconf
+_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
+
+
class RPCFail(Exception):
"""Class denoting RPC failure.
"""
+
def _Fail(msg, *args, **kwargs):
"""Log an error and the raise an RPCFail exception.
utils.RemoveFile(full_name)
+def _BuildUploadFileList():
+ """Build the list of allowed upload files.
+
+ This is abstracted so that it's built only once at module import time.
+
+ """
+ allowed_files = set([
+ constants.CLUSTER_CONF_FILE,
+ constants.ETC_HOSTS,
+ constants.SSH_KNOWN_HOSTS_FILE,
+ constants.VNC_PASSWORD_FILE,
+ constants.RAPI_CERT_FILE,
+ constants.RAPI_USERS_FILE,
+ constants.HMAC_CLUSTER_KEY,
+ ])
+
+ for hv_name in constants.HYPER_TYPES:
+ hv_class = hypervisor.GetHypervisorClass(hv_name)
+ allowed_files.update(hv_class.GetAncillaryFiles())
+
+ return frozenset(allowed_files)
+
+
+_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
+
+
def JobQueuePurge():
"""Removes job queue files and archived jobs.
master_node = cfg.GetMasterNode()
except errors.ConfigurationError, err:
_Fail("Cluster configuration incomplete: %s", err, exc=True)
- return master_netdev, master_ip, master_node
+ return (master_netdev, master_ip, master_node)
-def StartMaster(start_daemons):
+def StartMaster(start_daemons, no_voting):
"""Activate local node as master node.
The function will always try activate the IP address of the master
@type start_daemons: boolean
@param start_daemons: whether to also start the master
daemons (ganeti-masterd and ganeti-rapi)
+ @type no_voting: boolean
+ @param no_voting: whether to start ganeti-masterd without a node vote
+ (if start_daemons is True), but still non-interactively
@rtype: None
"""
# 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:
- msg = "Can't start daemon %s: %s" % (daemon, result.output)
- logging.error(msg)
- err_msgs.append(msg)
+ if no_voting:
+ masterd_args = "--no-voting --yes-do-it"
+ else:
+ masterd_args = ""
+
+ env = {
+ "EXTRA_MASTERD_ARGS": masterd_args,
+ }
+
+ result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
+ if result.failed:
+ msg = "Can't start Ganeti master: %s" % result.output
+ logging.error(msg)
+ err_msgs.append(msg)
if err_msgs:
_Fail("; ".join(err_msgs))
# 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)))
+ result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
+ if result.failed:
+ logging.error("Could not stop Ganeti master, command %s had exitcode %s"
+ " and error %s",
+ result.cmd, result.exit_code, result.output)
def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
utils.AddAuthorizedKey(auth_keys, sshpub)
- utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
+ result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
+ if result.failed:
+ _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
+ result.cmd, result.exit_code, result.output)
-def LeaveCluster():
+def LeaveCluster(modify_ssh_setup):
"""Cleans up and remove the current node.
This function cleans up and prepares the current node to be removed
L{errors.QuitGanetiException} which is used as a special case to
shutdown the node daemon.
+ @param modify_ssh_setup: boolean
+
"""
_CleanDirectory(constants.DATA_DIR)
JobQueuePurge()
- try:
- priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
-
- f = open(pub_key, 'r')
+ if modify_ssh_setup:
try:
- utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
- finally:
- f.close()
+ priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
- utils.RemoveFile(priv_key)
- utils.RemoveFile(pub_key)
- except errors.OpExecError:
- logging.exception("Error while processing ssh files")
+ utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
+
+ utils.RemoveFile(priv_key)
+ utils.RemoveFile(pub_key)
+ except errors.OpExecError:
+ logging.exception("Error while processing ssh files")
+
+ try:
+ utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
+ utils.RemoveFile(constants.RAPI_CERT_FILE)
+ utils.RemoveFile(constants.SSL_CERT_FILE)
+ except: # pylint: disable-msg=W0702
+ logging.exception("Error while removing cluster secrets")
+
+ result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
+ if result.failed:
+ logging.error("Command %s failed with exitcode %s and error %s",
+ result.cmd, result.exit_code, result.output)
# Raise a custom exception (handled in ganeti-noded)
raise errors.QuitGanetiException(True, 'Shutdown scheduled')
def GetNodeInfo(vgname, hypervisor_type):
- """Gives back a hash with different informations about the node.
+ """Gives back a hash with different information about the node.
@type vgname: C{string}
@param vgname: the name of the volume group to ask for disk space information
if hyp_info is not None:
outputarray.update(hyp_info)
- f = open("/proc/sys/kernel/random/boot_id", 'r')
- try:
- outputarray["bootid"] = f.read(128).rstrip("\n")
- finally:
- f.close()
+ outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
return outputarray
tmp[my_name] = ("Can't find my own primary/secondary IP"
" in the node list")
else:
- port = utils.GetNodeDaemonPort()
+ port = utils.GetDaemonPort(constants.NODED)
for name, pip, sip in what[constants.NV_NODENETTEST]:
fail = []
if not utils.TcpPing(pip, port, source=my_pip):
if constants.NV_VGLIST in what:
result[constants.NV_VGLIST] = utils.ListVolumeGroups()
+ if constants.NV_PVLIST in what:
+ result[constants.NV_PVLIST] = \
+ bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
+ filter_allocatable=False)
+
if constants.NV_VERSION in what:
result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
constants.RELEASE_VERSION)
used_minors = str(err)
result[constants.NV_DRBDLIST] = used_minors
+ if constants.NV_NODESETUP in what:
+ result[constants.NV_NODESETUP] = tmpr = []
+ if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
+ tmpr.append("The sysfs filesytem doesn't seem to be mounted"
+ " under /sys, missing required directories /sys/block"
+ " and /sys/class/net")
+ if (not os.path.isdir("/proc/sys") or
+ not os.path.isfile("/proc/sysrq-trigger")):
+ tmpr.append("The procfs filesystem doesn't seem to be mounted"
+ " under /proc, missing required directory /proc/sys and"
+ " the file /proc/sysrq-trigger")
+
+ if constants.NV_TIME in what:
+ result[constants.NV_TIME] = utils.SplitTime(time.time())
+
return result
name, size, attr = match.groups()
inactive = attr[4] == '-'
online = attr[5] == 'o'
+ virtual = attr[0] == 'v'
+ if virtual:
+ # we don't want to report such volumes as existing, since they
+ # don't really hold data
+ continue
lvs[name] = (size, inactive, online)
return lvs
missing.append(bridge)
if missing:
- _Fail("Missing bridges %s", ", ".join(missing))
+ _Fail("Missing bridges %s", utils.CommaJoin(missing))
def GetInstanceList(hypervisor_list):
def GetInstanceInfo(instance, hname):
- """Gives back the informations about an instance as a dictionary.
+ """Gives back the information about an instance as a dictionary.
@type instance: string
@param instance: the instance name
return output
-def InstanceOsAdd(instance, reinstall):
+def InstanceOsAdd(instance, reinstall, debug):
"""Add an OS to an instance.
@type instance: L{objects.Instance}
@param instance: Instance whose OS is to be installed
@type reinstall: boolean
@param reinstall: whether this is an instance reinstall
+ @type debug: integer
+ @param debug: debug level, passed to the OS scripts
@rtype: None
"""
inst_os = OSFromDisk(instance.os)
- create_env = OSEnvironment(instance)
+ create_env = OSEnvironment(instance, inst_os, debug)
if reinstall:
create_env['INSTANCE_REINSTALL'] = "1"
" log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
-def RunRenameInstance(instance, old_name):
+def RunRenameInstance(instance, old_name, debug):
"""Run the OS rename script for an instance.
@type instance: L{objects.Instance}
@param instance: Instance whose OS is to be installed
@type old_name: string
@param old_name: previous instance name
+ @type debug: integer
+ @param debug: debug level, passed to the OS scripts
@rtype: boolean
@return: the success of the operation
"""
inst_os = OSFromDisk(instance.os)
- rename_env = OSEnvironment(instance)
+ rename_env = OSEnvironment(instance, inst_os, debug)
rename_env['OLD_INSTANCE_NAME'] = old_name
logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
def _GetVGInfo(vg_name):
- """Get informations about the volume group.
+ """Get information about the volume group.
@type vg_name: str
@param vg_name: the volume group which we query
"vg_free": int(round(float(valarr[1]), 0)),
"pv_count": int(valarr[2]),
}
- except ValueError, err:
+ except (TypeError, ValueError), err:
logging.exception("Fail to parse vgs output: %s", err)
else:
logging.error("vgs output has the wrong number of fields (expected"
_Fail("Hypervisor error: %s", err, exc=True)
-def InstanceShutdown(instance):
+def InstanceShutdown(instance, timeout):
"""Shut an instance down.
@note: this functions uses polling with a hardcoded timeout.
@type instance: L{objects.Instance}
@param instance: the instance object
+ @type timeout: integer
+ @param timeout: maximum timeout for soft shutdown
@rtype: None
"""
hv_name = instance.hypervisor
- running_instances = GetInstanceList([hv_name])
+ hyper = hypervisor.GetHypervisor(hv_name)
iname = instance.name
- if iname not in running_instances:
+ if instance.name not in hyper.ListInstances():
logging.info("Instance %s not running, doing nothing", iname)
return
- hyper = hypervisor.GetHypervisor(hv_name)
- try:
- hyper.StopInstance(instance)
- except errors.HypervisorError, err:
- _Fail("Failed to stop instance %s: %s", iname, err)
+ class _TryShutdown:
+ def __init__(self):
+ self.tried_once = False
- # test every 10secs for 2min
+ def __call__(self):
+ if iname not in hyper.ListInstances():
+ return
- time.sleep(1)
- for dummy in range(11):
- if instance.name not in GetInstanceList([hv_name]):
- break
- time.sleep(10)
- else:
+ try:
+ hyper.StopInstance(instance, retry=self.tried_once)
+ except errors.HypervisorError, err:
+ if iname not in hyper.ListInstances():
+ # if the instance is no longer existing, consider this a
+ # success and go to cleanup
+ return
+
+ _Fail("Failed to stop instance %s: %s", iname, err)
+
+ self.tried_once = True
+
+ raise utils.RetryAgain()
+
+ try:
+ utils.Retry(_TryShutdown(), 5, timeout)
+ except utils.RetryTimeout:
# the shutdown did not succeed
- logging.error("Shutdown of '%s' unsuccessful, using destroy", iname)
+ logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
try:
hyper.StopInstance(instance, force=True)
except errors.HypervisorError, err:
- _Fail("Failed to force stop instance %s: %s", iname, err)
+ if iname in hyper.ListInstances():
+ # only raise an error if the instance still exists, otherwise
+ # the error could simply be "instance ... unknown"!
+ _Fail("Failed to force stop instance %s: %s", iname, err)
time.sleep(1)
- if instance.name in GetInstanceList([hv_name]):
+
+ if iname in hyper.ListInstances():
_Fail("Could not shutdown instance %s even by destroy", iname)
_RemoveBlockDevLinks(iname, instance.disks)
-def InstanceReboot(instance, reboot_type):
+def InstanceReboot(instance, reboot_type, shutdown_timeout):
"""Reboot an instance.
@type instance: L{objects.Instance}
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
+ - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
+ not accepted here, since that mode is handled differently, in
+ cmdlib, and translates into full stop and start of the
+ instance (instead of a call_instance_reboot RPC)
+ @type shutdown_timeout: integer
+ @param shutdown_timeout: maximum timeout for soft shutdown
@rtype: None
"""
_Fail("Failed to soft reboot instance %s: %s", instance.name, err)
elif reboot_type == constants.INSTANCE_REBOOT_HARD:
try:
- InstanceShutdown(instance)
+ InstanceShutdown(instance, shutdown_timeout)
return StartInstance(instance)
except errors.HypervisorError, err:
_Fail("Failed to hard reboot instance %s: %s", instance.name, err)
hyper = hypervisor.GetHypervisor(instance.hypervisor)
try:
- hyper.MigrateInstance(instance.name, target, live)
+ hyper.MigrateInstance(instance, target, live)
except errors.HypervisorError, err:
_Fail("Failed to migrate instance: %s", err, exc=True)
it's not required to return anything.
"""
+ # TODO: remove the obsolete 'size' argument
+ # pylint: disable-msg=W0613
clist = []
if disk.children:
for child in disk.children:
# we need the children open in case the device itself has to
# be assembled
try:
+ # pylint: disable-msg=E1103
crdev.Open()
except errors.BlockDeviceError, err:
_Fail("Can't make child '%s' read-write: %s", child, err)
try:
result = _RecursiveAssembleBD(disk, owner, as_primary)
if isinstance(result, bdev.BlockDev):
+ # pylint: disable-msg=E1103
result = result.dev_path
except errors.BlockDeviceError, err:
_Fail("Error while assembling disk: %s", err, exc=True)
def BlockdevShutdown(disk):
"""Shut down a block device.
- First, if the device is assembled (Attach() is successfull), then
+ First, if the device is assembled (Attach() is successful), then
the device is shutdown. Then the children of the device are
shutdown.
rbd = _RecursiveFindBD(dsk)
if rbd is None:
_Fail("Can't find device %s", dsk)
+
stats.append(rbd.CombinedSyncStatus())
+
return stats
def _RecursiveFindBD(disk):
"""Check if a device is activated.
- If so, return informations about the real device.
+ If so, return information about the real device.
@type disk: L{objects.Disk}
@param disk: the disk object we need to find
def BlockdevFind(disk):
"""Check if a device is activated.
- If it is, return informations about the real device.
+ If it is, return information about the real device.
@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)
+ @rtype: None or objects.BlockDevStatus
+ @return: None if the disk cannot be found, otherwise a the current
+ information
"""
try:
rbd = _RecursiveFindBD(disk)
except errors.BlockDeviceError, err:
_Fail("Failed to find device: %s", err, exc=True)
+
if rbd is None:
return None
- return (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus()
+
+ return rbd.GetSyncStatus()
+
+
+def BlockdevGetsize(disks):
+ """Computes the size of the given disks.
+
+ If a disk is not found, returns None instead.
+
+ @type disks: list of L{objects.Disk}
+ @param disks: the list of disk to compute the size for
+ @rtype: list
+ @return: list with elements None if the disk cannot be found,
+ otherwise the size
+
+ """
+ result = []
+ for cf in disks:
+ try:
+ rbd = _RecursiveFindBD(cf)
+ except errors.BlockDeviceError:
+ result.append(None)
+ continue
+ if rbd is None:
+ result.append(None)
+ else:
+ result.append(rbd.GetActualSize())
+ return result
+
+
+def BlockdevExport(disk, dest_node, dest_path, cluster_name):
+ """Export a block device to a remote node.
+
+ @type disk: L{objects.Disk}
+ @param disk: the description of the disk to export
+ @type dest_node: str
+ @param dest_node: the destination node to export to
+ @type dest_path: str
+ @param dest_path: the destination path on the target node
+ @type cluster_name: str
+ @param cluster_name: the cluster name, needed for SSH hostalias
+ @rtype: None
+
+ """
+ real_disk = _RecursiveFindBD(disk)
+ if real_disk is None:
+ _Fail("Block device '%s' is not set up", disk)
+
+ real_disk.Open()
+
+ # the block size on the read dd is 1MiB to match our units
+ expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
+ "dd if=%s bs=1048576 count=%s",
+ real_disk.dev_path, str(disk.size))
+
+ # we set here a smaller block size as, due to ssh buffering, more
+ # than 64-128k will mostly ignored; we use nocreat to fail if the
+ # device is not already there or we pass a wrong path; we use
+ # notrunc to no attempt truncate on an LV device; we use oflag=dsync
+ # to not buffer too much memory; this means that at best, we flush
+ # every 64k, which will not be very fast
+ destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
+ " oflag=dsync", dest_path)
+
+ 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, utils.ShellQuoteArgs(remotecmd)])
+
+ result = utils.RunCmd(["bash", "-c", command])
+
+ if result.failed:
+ _Fail("Disk copy command '%s' returned error: %s"
+ " output: %s", command, result.fail_reason, result.output)
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
if not os.path.isabs(file_name):
_Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
- allowed_files = set([
- constants.CLUSTER_CONF_FILE,
- constants.ETC_HOSTS,
- constants.SSH_KNOWN_HOSTS_FILE,
- constants.VNC_PASSWORD_FILE,
- constants.RAPI_CERT_FILE,
- constants.RAPI_USERS_FILE,
- ])
-
- for hv_name in constants.HYPER_TYPES:
- hv_class = hypervisor.GetHypervisor(hv_name)
- allowed_files.update(hv_class.GetAncillaryFiles())
-
- if file_name not in allowed_files:
+ if file_name not in _ALLOWED_UPLOAD_FILES:
_Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
file_name)
return detail
-def _OSOndiskAPIVersion(name, os_dir):
+def _OSOndiskAPIVersion(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
- the 'name' parameter and residing in the 'os_dir' directory.
+ This function will try to read the API version of the OS residing in
+ the 'os_dir' directory.
- @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
+ @param os_dir: the directory in which we should look for the OS
@rtype: tuple
@return: tuple (status, data) with status denoting the validity and
data holding either the vaid versions or an error message
"""
- api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
+ api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
try:
st = os.stat(api_file)
except EnvironmentError, err:
- return False, ("Required file 'ganeti_api_version' file not"
- " found under path %s: %s" % (os_dir, _ErrnoOrStr(err)))
+ return False, ("Required file '%s' not found under path %s: %s" %
+ (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
- return False, ("File 'ganeti_api_version' file at %s is not"
- " a regular file" % os_dir)
+ return False, ("File '%s' in %s is not a regular file" %
+ (constants.OS_API_FILE, os_dir))
try:
- f = open(api_file)
- try:
- api_versions = f.readlines()
- finally:
- f.close()
+ api_versions = utils.ReadFile(api_file).splitlines()
except EnvironmentError, err:
return False, ("Error while reading the API version file at %s: %s" %
(api_file, _ErrnoOrStr(err)))
- api_versions = [version.strip() for version in api_versions]
try:
- api_versions = [int(version) for version in api_versions]
+ api_versions = [int(version.strip()) for version in api_versions]
except (TypeError, ValueError), err:
return False, ("API version(s) can't be converted to integer: %s" %
str(err))
search (if not given defaults to
L{constants.OS_SEARCH_PATH})
@rtype: list of L{objects.OS}
- @return: a list of tuples (name, path, status, diagnose)
+ @return: a list of tuples (name, path, status, diagnose, variants)
for all (potential) OSes under all search paths, where:
- name is the (potential) OS name
- path is the full path to the OS
- status True/False is the validity of the OS
- diagnose is the error message for an invalid OS, otherwise empty
+ - variants is a list of supported OS variants, if any
"""
if top_dirs is None:
status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
if status:
diagnose = ""
+ variants = os_inst.supported_variants
else:
diagnose = os_inst
- result.append((name, os_path, status, diagnose))
+ variants = []
+ result.append((name, os_path, status, diagnose, variants))
return result
"""
if base_dir is None:
os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
- if os_dir is None:
- return False, "Directory for OS %s not found in search path" % name
else:
- os_dir = os.path.sep.join([base_dir, name])
+ os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
+
+ if os_dir is None:
+ return False, "Directory for OS %s not found in search path" % name
- status, api_versions = _OSOndiskAPIVersion(name, os_dir)
+ status, api_versions = _OSOndiskAPIVersion(os_dir)
if not status:
# push the error up
return status, api_versions
- if constants.OS_API_VERSION not in api_versions:
+ if not constants.OS_API_VERSIONS.intersection(api_versions):
return False, ("API version mismatch for path '%s': found %s, want %s." %
- (os_dir, api_versions, constants.OS_API_VERSION))
+ (os_dir, api_versions, constants.OS_API_VERSIONS))
- # OS Scripts dictionary, we will populate it with the actual script names
- os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
+ # OS Files dictionary, we will populate it with the absolute path names
+ os_files = dict.fromkeys(constants.OS_SCRIPTS)
- for script in os_scripts:
- os_scripts[script] = os.path.sep.join([os_dir, script])
+ if max(api_versions) >= constants.OS_API_V15:
+ os_files[constants.OS_VARIANTS_FILE] = ''
+
+ for filename in os_files:
+ os_files[filename] = os.path.sep.join([os_dir, filename])
try:
- st = os.stat(os_scripts[script])
+ st = os.stat(os_files[filename])
except EnvironmentError, err:
- return False, ("Script '%s' under path '%s' is missing (%s)" %
- (script, os_dir, _ErrnoOrStr(err)))
-
- if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
- return False, ("Script '%s' under path '%s' is not executable" %
- (script, os_dir))
+ return False, ("File '%s' under path '%s' is missing (%s)" %
+ (filename, os_dir, _ErrnoOrStr(err)))
if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
- return False, ("Script '%s' under path '%s' is not a regular file" %
- (script, os_dir))
+ return False, ("File '%s' under path '%s' is not a regular file" %
+ (filename, os_dir))
+
+ if filename in constants.OS_SCRIPTS:
+ if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
+ return False, ("File '%s' under path '%s' is not executable" %
+ (filename, os_dir))
+
+ variants = None
+ if constants.OS_VARIANTS_FILE in os_files:
+ variants_file = os_files[constants.OS_VARIANTS_FILE]
+ try:
+ variants = utils.ReadFile(variants_file).splitlines()
+ except EnvironmentError, err:
+ return False, ("Error while reading the OS variants file at %s: %s" %
+ (variants_file, _ErrnoOrStr(err)))
+ if not variants:
+ return False, ("No supported os variant found")
os_obj = objects.OS(name=name, path=os_dir,
- 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],
+ create_script=os_files[constants.OS_SCRIPT_CREATE],
+ export_script=os_files[constants.OS_SCRIPT_EXPORT],
+ import_script=os_files[constants.OS_SCRIPT_IMPORT],
+ rename_script=os_files[constants.OS_SCRIPT_RENAME],
+ supported_variants=variants,
api_versions=api_versions)
return True, os_obj
@raise RPCFail: if we don't find a valid OS
"""
- status, payload = _TryOSFromDisk(name, base_dir)
+ name_only = name.split("+", 1)[0]
+ status, payload = _TryOSFromDisk(name_only, base_dir)
if not status:
_Fail(payload)
return payload
-def OSEnvironment(instance, debug=0):
+def OSEnvironment(instance, inst_os, debug=0):
"""Calculate the environment for an os script.
@type instance: L{objects.Instance}
@param instance: target instance for the os script run
+ @type inst_os: L{objects.OS}
+ @param inst_os: operating system for which the environment is being built
@type debug: integer
@param debug: debug level (0 or 1, for OS Api 10)
@rtype: dict
"""
result = {}
- result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
+ api_version = \
+ max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
+ result['OS_API_VERSION'] = '%d' % 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)
result['DEBUG_LEVEL'] = '%d' % debug
+ if api_version >= constants.OS_API_V15:
+ try:
+ variant = instance.os.split('+', 1)[1]
+ except IndexError:
+ variant = inst_os.supported_variants[0]
+ result['OS_VARIANT'] = variant
for idx, disk in enumerate(instance.disks):
real_disk = _RecursiveFindBD(disk)
if real_disk is None:
@return: snapshot disk path
"""
- if disk.children:
- if len(disk.children) == 1:
- # only one child, let's recurse on it
- 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 BlockdevSnapshot(child)
+ if disk.dev_type == constants.LD_DRBD8:
+ if not disk.children:
+ _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
+ disk.unique_id)
+ return BlockdevSnapshot(disk.children[0])
elif disk.dev_type == constants.LD_LV:
r_dev = _RecursiveFindBD(disk)
if r_dev is not None:
+ # FIXME: choose a saner value for the snapshot size
# let's stay on the safe side and ask for the full size, for now
return r_dev.Snapshot(disk.size)
else:
disk.unique_id, disk.dev_type)
-def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
+def ExportSnapshot(disk, dest_node, instance, cluster_name, idx, debug):
"""Export a block device snapshot to a remote node.
@type disk: L{objects.Disk}
@type idx: int
@param idx: the index of the disk in the instance's disk list,
used to export to the OS scripts environment
+ @type debug: integer
+ @param debug: debug level, passed to the OS scripts
@rtype: None
"""
- export_env = OSEnvironment(instance)
-
inst_os = OSFromDisk(instance.os)
+ export_env = OSEnvironment(instance, inst_os, debug)
+
export_script = inst_os.export_script
logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
# 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 2>%s", inst_os.path,
- export_script, logfile)
+ expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
+ inst_os.path, export_script, logfile)
comprcmd = "gzip"
# all commands have been checked, so we're safe to combine them
command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
- result = utils.RunCmd(command, env=export_env)
+ result = utils.RunCmd(["bash", "-c", command], env=export_env)
if result.failed:
_Fail("OS snapshot export command '%s' returned error: %s"
return config.Dumps()
-def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
+def ImportOSIntoInstance(instance, src_node, src_images, cluster_name, debug):
"""Import an os image into an instance.
@type instance: L{objects.Instance}
@param src_node: source node for the disk images
@type src_images: list of string
@param src_images: absolute paths of the disk images
+ @type debug: integer
+ @param debug: debug level, passed to the OS scripts
@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_env = OSEnvironment(instance, inst_os, debug)
import_script = inst_os.import_script
logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
@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
+ whether the operation was successful
"""
file_storage_dir = _TransformFileStorageDir(file_storage_dir)
master, myself = ssconf.GetMasterAndMyself()
if master == myself:
_Fail("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)):
+
+ result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
+ if not result.failed:
_Fail("The master daemon is running, will not demote")
+
try:
- utils.CreateBackup(constants.CLUSTER_CONF_FILE)
+ if os.path.isfile(constants.CLUSTER_CONF_FILE):
+ utils.CreateBackup(constants.CLUSTER_CONF_FILE)
except EnvironmentError, err:
if err.errno != errno.ENOENT:
_Fail("Error while backing up cluster file: %s", err, exc=True)
+
utils.RemoveFile(constants.CLUSTER_CONF_FILE)
rd.AttachNet(multimaster)
except errors.BlockDeviceError, err:
_Fail("Can't change network configuration: %s", err)
+
# wait until the disks are connected; we need to retry the re-attach
# if the device becomes standalone, as this might happen if the one
# node disconnects and reconnects in a different mode before the
# other node reconnects; in this case, one or both of the nodes will
# decide it has wrong configuration and switch to standalone
- RECONNECT_TIMEOUT = 2 * 60
- sleep_time = 0.100 # start with 100 miliseconds
- timeout_limit = time.time() + RECONNECT_TIMEOUT
- while time.time() < timeout_limit:
+
+ def _Attach():
all_connected = True
+
for rd in bdevs:
stats = rd.GetProcStatus()
- if not (stats.is_connected or stats.is_in_resync):
- all_connected = False
+
+ all_connected = (all_connected and
+ (stats.is_connected or stats.is_in_resync))
+
if stats.is_standalone:
# peer had different config info and this node became
# standalone, even though this should not happen with the
rd.AttachNet(multimaster)
except errors.BlockDeviceError, err:
_Fail("Can't change network configuration: %s", err)
- if all_connected:
- break
- time.sleep(sleep_time)
- sleep_time = min(5, sleep_time * 1.5)
- if not all_connected:
+
+ if not all_connected:
+ raise utils.RetryAgain()
+
+ try:
+ # Start with a delay of 100 miliseconds and go up to 5 seconds
+ utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
+ except utils.RetryTimeout:
_Fail("Timeout in disk reconnecting")
+
if multimaster:
# change to primary mode
for rd in bdevs:
"""Wait until DRBDs have synchronized.
"""
+ def _helper(rd):
+ stats = rd.GetProcStatus()
+ if not (stats.is_connected or stats.is_in_resync):
+ raise utils.RetryAgain()
+ return stats
+
bdevs = _FindDisks(nodes_ip, disks)
min_resync = 100
alldone = True
for rd in bdevs:
- stats = rd.GetProcStatus()
- if not (stats.is_connected or stats.is_in_resync):
- _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
+ try:
+ # poll each second for 15 seconds
+ stats = utils.Retry(_helper, 1, 15, args=[rd])
+ except utils.RetryTimeout:
+ stats = rd.GetProcStatus()
+ # last check
+ if not (stats.is_connected or stats.is_in_resync):
+ _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
alldone = alldone and (not stats.is_in_resync)
if stats.sync_percent is not None:
min_resync = min(min_resync, stats.sync_percent)
on the master side.
"""
- RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
-
def __init__(self, hooks_base_dir=None):
"""Constructor for hooks runner.
"""
if hooks_base_dir is None:
hooks_base_dir = constants.HOOKS_BASE_DIR
- self._BASE_DIR = hooks_base_dir
+ # yeah, _BASE_DIR is not valid for attributes, we use it like a
+ # constant
+ self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
@staticmethod
def ExecHook(script, env):
for relname in dir_contents:
fname = os.path.join(dir_name, relname)
if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
- self.RE_MASK.match(relname) is not None):
+ constants.EXT_PLUGIN_MASK.match(relname) is not None):
rrval = constants.HKR_SKIP
output = ""
else:
the master side.
"""
- def Run(self, name, idata):
+ @staticmethod
+ def Run(name, idata):
"""Run an iallocator script.
@type name: str