X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/7c0aa8e9dac30a6f2f17dd60c59c4cccf85cac2a..bdefe5dd311f72b591a4dcf50e01715930d2f00d:/lib/backend.py?ds=sidebyside diff --git a/lib/backend.py b/lib/backend.py index 6122acd..367e89c 100644 --- a/lib/backend.py +++ b/lib/backend.py @@ -23,9 +23,17 @@ @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in the L{UploadFile} function +@var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted + in the L{_CleanDirectory} 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 import os.path @@ -34,7 +42,6 @@ import time import stat import errno import re -import subprocess import random import logging import tempfile @@ -52,6 +59,11 @@ from ganeti import ssconf _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id" +_ALLOWED_CLEAN_DIRS = frozenset([ + constants.DATA_DIR, + constants.JOB_QUEUE_ARCHIVE_DIR, + constants.QUEUE_DIR, + ]) class RPCFail(Exception): @@ -138,6 +150,10 @@ def _CleanDirectory(path, exclude=None): to the empty list """ + if path not in _ALLOWED_CLEAN_DIRS: + _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'", + path) + if not os.path.isdir(path): return if exclude is None: @@ -147,7 +163,7 @@ def _CleanDirectory(path, exclude=None): 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)) + full_name = utils.PathJoin(path, rel_name) if full_name in exclude: continue if os.path.isfile(full_name) and not os.path.islink(full_name): @@ -255,21 +271,20 @@ def StartMaster(start_daemons, no_voting): # and now start the master and rapi daemons if start_daemons: - daemons_params = { - 'ganeti-masterd': [], - 'ganeti-rapi': [], - } if no_voting: - daemons_params['ganeti-masterd'].append('--no-voting') - daemons_params['ganeti-masterd'].append('--yes-do-it') - for daemon in daemons_params: - cmd = [daemon] - cmd.extend(daemons_params[daemon]) - result = utils.RunCmd(cmd) - if result.failed: - msg = "Can't start daemon %s: %s" % (daemon, result.output) - logging.error(msg) - err_msgs.append(msg) + 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)) @@ -301,9 +316,11 @@ def StopMaster(stop_daemons): # but otherwise ignore the failure if stop_daemons: - # stop/kill the rapi and the master daemon - for daemon in constants.RAPI, constants.MASTERD: - 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): @@ -348,10 +365,13 @@ 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 @@ -361,31 +381,34 @@ def LeaveCluster(): 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) + if modify_ssh_setup: + try: + priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) - utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) + 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") + 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: + except: # pylint: disable-msg=W0702 logging.exception("Error while removing cluster secrets") - confd_pid = utils.ReadPidFile(utils.DaemonPidFileName(constants.CONFD)) - - if confd_pid: - utils.KillProcess(confd_pid, timeout=2) + 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') @@ -506,6 +529,11 @@ def VerifyNode(what, cluster_name): 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) @@ -533,6 +561,10 @@ def VerifyNode(what, cluster_name): 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 @@ -648,7 +680,7 @@ def BridgesExist(bridges_list): missing.append(bridge) if missing: - _Fail("Missing bridges %s", ", ".join(missing)) + _Fail("Missing bridges %s", utils.CommaJoin(missing)) def GetInstanceList(hypervisor_list): @@ -767,24 +799,43 @@ def GetAllInstancesInfo(hypervisor_list): return output -def InstanceOsAdd(instance, reinstall): +def _InstanceLogName(kind, os_name, instance): + """Compute the OS log filename for a given instance and operation. + + The instance name and os name are passed in as strings since not all + operations have these as part of an instance object. + + @type kind: string + @param kind: the operation type (e.g. add, import, etc.) + @type os_name: string + @param os_name: the os name + @type instance: string + @param instance: the name of the instance being imported/added/etc. + + """ + base = "%s-%s-%s-%d.log" % (kind, os_name, instance, int(time.time())) + return utils.PathJoin(constants.LOG_OS_DIR, base) + + +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, inst_os) + create_env = OSEnvironment(instance, inst_os, debug) if reinstall: create_env['INSTANCE_REINSTALL'] = "1" - logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os, - instance.name, int(time.time())) + logfile = _InstanceLogName("add", instance.os, instance.name) result = utils.RunCmd([inst_os.create_script], env=create_env, cwd=inst_os.path, output=logfile,) @@ -798,25 +849,26 @@ def InstanceOsAdd(instance, reinstall): " 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, inst_os) + 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, - old_name, - instance.name, int(time.time())) + logfile = _InstanceLogName("rename", instance.os, + "%s-%s" % (old_name, instance.name)) result = utils.RunCmd([inst_os.rename_script], env=rename_env, cwd=inst_os.path, output=logfile) @@ -862,7 +914,7 @@ def _GetVGInfo(vg_name): "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" @@ -871,8 +923,8 @@ def _GetVGInfo(vg_name): def _GetBlockDevSymlinkPath(instance_name, idx): - return os.path.join(constants.DISK_LINKS_DIR, - "%s:%d" % (instance_name, idx)) + return utils.PathJoin(constants.DISK_LINKS_DIR, + "%s:%d" % (instance_name, idx)) def _SymlinkBlockDev(instance_name, device_path, idx): @@ -985,42 +1037,51 @@ def InstanceShutdown(instance, timeout): """ hv_name = instance.hypervisor hyper = hypervisor.GetHypervisor(hv_name) - running_instances = hyper.ListInstances() 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 - start = time.time() - end = start + timeout - sleep_time = 1 + class _TryShutdown: + def __init__(self): + self.tried_once = False - tried_once = False - while not tried_once and time.time() < end: - try: - hyper.StopInstance(instance, retry=tried_once) - except errors.HypervisorError, err: - _Fail("Failed to stop instance %s: %s", iname, err) - tried_once = True - time.sleep(sleep_time) - if instance.name not in hyper.ListInstances(): - break - if sleep_time < 5: - # 1.2 behaves particularly good for our case: - # it gives us 10 increasing steps and caps just slightly above 5 seconds - sleep_time *= 1.2 - else: + def __call__(self): + if iname not in hyper.ListInstances(): + return + + 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, 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) @@ -1042,8 +1103,8 @@ def InstanceReboot(instance, reboot_type, shutdown_timeout): 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 timeout: integer - @param timeout: maximum timeout for soft shutdown + @type shutdown_timeout: integer + @param shutdown_timeout: maximum timeout for soft shutdown @rtype: None """ @@ -1138,7 +1199,7 @@ def MigrateInstance(instance, target, live): 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) @@ -1164,6 +1225,8 @@ def BlockdevCreate(disk, size, owner, on_primary, info): 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: @@ -1175,6 +1238,7 @@ def BlockdevCreate(disk, size, owner, on_primary, info): # 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) @@ -1309,6 +1373,7 @@ def BlockdevAssemble(disk, owner, as_primary): 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) @@ -1396,6 +1461,8 @@ def BlockdevRemovechildren(parent_cdev, new_cdevs): else: devs.append(bd.dev_path) else: + if not utils.IsNormAbsPath(rpath): + _Fail("Strange path returned from StaticDevPath: '%s'", rpath) devs.append(rpath) parent_bdev.RemoveChildren(devs) @@ -1483,7 +1550,7 @@ def BlockdevGetsize(disks): for cf in disks: try: rbd = _RecursiveFindBD(cf) - except errors.BlockDeviceError, err: + except errors.BlockDeviceError: result.append(None) continue if rbd is None: @@ -1604,22 +1671,20 @@ def _ErrnoOrStr(err): 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, constants.OS_API_FILE]) + api_file = utils.PathJoin(os_dir, constants.OS_API_FILE) try: st = os.stat(api_file) @@ -1675,7 +1740,7 @@ def DiagnoseOS(top_dirs=None): logging.exception("Can't list the OS directory %s: %s", dir_name, err) break for name in f_names: - os_path = os.path.sep.join([dir_name, name]) + os_path = utils.PathJoin(dir_name, name) status, os_inst = _TryOSFromDisk(name, base_dir=dir_name) if status: diagnose = "" @@ -1704,12 +1769,13 @@ def _TryOSFromDisk(name, base_dir=None): """ 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) - status, api_versions = _OSOndiskAPIVersion(name, os_dir) + if os_dir is None: + return False, "Directory for OS %s not found in search path" % name + + status, api_versions = _OSOndiskAPIVersion(os_dir) if not status: # push the error up return status, api_versions @@ -1724,23 +1790,23 @@ def _TryOSFromDisk(name, base_dir=None): if max(api_versions) >= constants.OS_API_V15: os_files[constants.OS_VARIANTS_FILE] = '' - for name in os_files: - os_files[name] = os.path.sep.join([os_dir, name]) + for filename in os_files: + os_files[filename] = utils.PathJoin(os_dir, filename) try: - st = os.stat(os_files[name]) + st = os.stat(os_files[filename]) except EnvironmentError, err: return False, ("File '%s' under path '%s' is missing (%s)" % - (name, os_dir, _ErrnoOrStr(err))) + (filename, os_dir, _ErrnoOrStr(err))) if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): return False, ("File '%s' under path '%s' is not a regular file" % - (name, os_dir)) + (filename, os_dir)) - if name in constants.OS_SCRIPTS: + 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" % - (name, os_dir)) + (filename, os_dir)) variants = None if constants.OS_VARIANTS_FILE in os_files: @@ -1790,13 +1856,13 @@ def OSFromDisk(name, base_dir=None): return payload -def OSEnvironment(instance, os, 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 os: L{objects.OS} - @param os: operating system for which the environment is being built + @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 @@ -1806,7 +1872,8 @@ def OSEnvironment(instance, os, debug=0): """ result = {} - api_version = max(constants.OS_API_VERSIONS.intersection(os.api_versions)) + 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 @@ -1818,7 +1885,7 @@ def OSEnvironment(instance, os, debug=0): try: variant = instance.os.split('+', 1)[1] except IndexError: - variant = os.supported_variants[0] + variant = inst_os.supported_variants[0] result['OS_VARIANT'] = variant for idx, disk in enumerate(instance.disks): real_disk = _RecursiveFindBD(disk) @@ -1891,19 +1958,15 @@ def BlockdevSnapshot(disk): @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: @@ -1913,7 +1976,7 @@ def BlockdevSnapshot(disk): 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} @@ -1927,16 +1990,17 @@ def ExportSnapshot(disk, dest_node, instance, cluster_name, idx): @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 """ inst_os = OSFromDisk(instance.os) - export_env = OSEnvironment(instance, inst_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, - instance.name, int(time.time())) + logfile = _InstanceLogName("export", inst_os.name, instance.name) if not os.path.exists(constants.LOG_OS_DIR): os.mkdir(constants.LOG_OS_DIR, 0750) real_disk = _RecursiveFindBD(disk) @@ -1948,7 +2012,7 @@ def ExportSnapshot(disk, dest_node, instance, cluster_name, idx): export_env['EXPORT_DEVICE'] = real_disk.dev_path export_env['EXPORT_INDEX'] = str(idx) - destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new") + destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new") destfile = disk.physical_id[1] # the target command is built out of three individual commands, @@ -1959,8 +2023,8 @@ def ExportSnapshot(disk, dest_node, instance, cluster_name, idx): comprcmd = "gzip" - destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s", - destdir, destdir, destfile) + destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s", + destdir, utils.PathJoin(destdir, destfile)) remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node, constants.GANETI_RUNAS, destcmd) @@ -1988,8 +2052,8 @@ def FinalizeExport(instance, snap_disks): @rtype: None """ - destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new") - finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name) + destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new") + finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name) config = objects.SerializableConfigParser() @@ -2032,7 +2096,7 @@ def FinalizeExport(instance, snap_disks): config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total) - utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE), + utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE), data=config.Dumps()) shutil.rmtree(finaldestdir, True) shutil.move(destdir, finaldestdir) @@ -2049,7 +2113,7 @@ def ExportInfo(dest): export info """ - cff = os.path.join(dest, constants.EXPORT_CONF_FILE) + cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE) config = objects.SerializableConfigParser() config.read(cff) @@ -2061,7 +2125,7 @@ def ExportInfo(dest): 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} @@ -2070,16 +2134,17 @@ def ImportOSIntoInstance(instance, src_node, src_images, cluster_name): @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 """ inst_os = OSFromDisk(instance.os) - import_env = OSEnvironment(instance, inst_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, - instance.name, int(time.time())) + logfile = _InstanceLogName("import", instance.os, instance.name) if not os.path.exists(constants.LOG_OS_DIR): os.mkdir(constants.LOG_OS_DIR, 0750) @@ -2130,7 +2195,7 @@ def RemoveExport(export): @rtype: None """ - target = os.path.join(constants.EXPORT_DIR, export) + target = utils.PathJoin(constants.EXPORT_DIR, export) try: shutil.rmtree(target) @@ -2192,6 +2257,8 @@ def _TransformFileStorageDir(file_storage_dir): @return: the normalized path if valid, None otherwise """ + if not constants.ENABLE_FILE_STORAGE: + _Fail("File storage disabled at configure time") cfg = _GetConfig() file_storage_dir = os.path.normpath(file_storage_dir) base_file_storage_dir = cfg.GetFileStorageDir() @@ -2417,15 +2484,18 @@ def DemoteFromMC(): 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) - 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: 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) @@ -2482,20 +2552,22 @@ def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster): 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 @@ -2504,12 +2576,16 @@ def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster): 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: @@ -2523,14 +2599,25 @@ def DrbdWaitSync(nodes_ip, disks): """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) @@ -2564,8 +2651,6 @@ class HooksRunner(object): on the master side. """ - RE_MASK = re.compile("^[a-zA-Z0-9_-]+$") - def __init__(self, hooks_base_dir=None): """Constructor for hooks runner. @@ -2576,57 +2661,9 @@ class HooksRunner(object): """ if hooks_base_dir is None: hooks_base_dir = constants.HOOKS_BASE_DIR - self._BASE_DIR = hooks_base_dir - - @staticmethod - def ExecHook(script, env): - """Exec one hook 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 - fdstdin = None - try: - fdstdin = open("/dev/null", "r") - child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, close_fds=True, - shell=False, cwd="/", env=env) - output = "" - try: - output = child.stdout.read(4096) - child.stdout.close() - except EnvironmentError, err: - output += "Hook script error: %s" % str(err) - - while True: - try: - result = child.wait() - break - except EnvironmentError, err: - if err.errno == errno.EINTR: - continue - raise - finally: - # try not to leak fds - for fd in (fdstdin, ): - if fd is not None: - try: - fd.close() - except EnvironmentError, err: - # just log the error - #logging.exception("Error while closing fd %s", fd) - pass - - return result == 0, utils.SafeEncode(output.strip()) + # yeah, _BASE_DIR is not valid for attributes, we use it like a + # constant + self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103 def RunHooks(self, hpath, phase, env): """Run the scripts in the hooks directory. @@ -2657,34 +2694,35 @@ class HooksRunner(object): else: _Fail("Unknown hooks phase '%s'", phase) - rr = [] subdir = "%s-%s.d" % (hpath, suffix) - dir_name = "%s/%s" % (self._BASE_DIR, subdir) - try: - dir_contents = utils.ListVisibleFiles(dir_name) - except OSError: - # FIXME: must log output in case of failures - return rr - - # we use the standard python sort order, - # so 00name is the recommended naming scheme - dir_contents.sort() - 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): + dir_name = utils.PathJoin(self._BASE_DIR, subdir) + + results = [] + + if not os.path.isdir(dir_name): + # for non-existing/non-dirs, we simply exit instead of logging a + # warning at every operation + return results + + runparts_results = utils.RunParts(dir_name, env=env, reset_env=True) + + for (relname, relstatus, runresult) in runparts_results: + if relstatus == constants.RUNPARTS_SKIP: rrval = constants.HKR_SKIP output = "" - else: - result, output = self.ExecHook(fname, env) - if not result: + elif relstatus == constants.RUNPARTS_ERR: + rrval = constants.HKR_FAIL + output = "Hook script execution error: %s" % runresult + elif relstatus == constants.RUNPARTS_RUN: + if runresult.failed: rrval = constants.HKR_FAIL else: rrval = constants.HKR_SUCCESS - rr.append(("%s/%s" % (subdir, relname), rrval, output)) + output = utils.SafeEncode(runresult.output.strip()) + results.append(("%s/%s" % (subdir, relname), rrval, output)) - return rr + return results class IAllocatorRunner(object): @@ -2694,7 +2732,8 @@ class IAllocatorRunner(object): the master side. """ - def Run(self, name, idata): + @staticmethod + def Run(name, idata): """Run an iallocator script. @type name: str @@ -2750,7 +2789,7 @@ class DevCacheManager(object): if dev_path.startswith(cls._DEV_PREFIX): dev_path = dev_path[len(cls._DEV_PREFIX):] dev_path = dev_path.replace("/", "_") - fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path) + fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path) return fpath @classmethod