X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/007a2f3e50e20aa187eff32807c221afb0227986..8c229cc7e24f1c7075b38ff0e21aab7c75299c4d:/lib/backend.py?ds=sidebyside diff --git a/lib/backend.py b/lib/backend.py index d2c154a..382eec5 100644 --- a/lib/backend.py +++ b/lib/backend.py @@ -26,11 +26,11 @@ import os import os.path import shutil import time -import tempfile import stat import errno import re import subprocess +import random from ganeti import logger from ganeti import errors @@ -43,6 +43,10 @@ from ganeti import objects from ganeti import ssconf +def _GetSshRunner(): + return ssh.SshRunner() + + def StartMaster(): """Activate local node as master node. @@ -64,9 +68,7 @@ def StartMaster(): def StopMaster(): """Deactivate this node as master. - This does two things: - - run the master stop script - - remove link to master cron script. + This runs the master stop script. """ result = utils.RunCmd([constants.MASTER_SCRIPT, "-d", "stop"]) @@ -122,7 +124,6 @@ def LeaveCluster(): if os.path.isfile(full_name) and not os.path.islink(full_name): utils.RemoveFile(full_name) - try: priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) except errors.OpExecError, err: @@ -138,6 +139,9 @@ def LeaveCluster(): 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): """Gives back a hash with different informations about the node. @@ -200,10 +204,38 @@ def VerifyNode(what): if 'nodelist' in what: result['nodelist'] = {} + random.shuffle(what['nodelist']) for node in what['nodelist']: - success, message = ssh.VerifyNodeHostname(node) + success, message = _GetSshRunner().VerifyNodeHostname(node) if not success: result['nodelist'][node] = message + if 'node-net-test' in what: + result['node-net-test'] = {} + my_name = utils.HostInfo().name + my_pip = my_sip = None + for name, pip, sip in what['node-net-test']: + if name == my_name: + my_pip = pip + my_sip = sip + break + if not my_pip: + result['node-net-test'][my_name] = ("Can't find my own" + " primary/secondary IP" + " in the node list") + else: + port = ssconf.SimpleStore().GetNodeDaemonPort() + for name, pip, sip in what['node-net-test']: + 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: + result['node-net-test'][name] = ("failure using the %s" + " interface(s)" % + " and ".join(fail)) + return result @@ -211,19 +243,31 @@ def GetVolumeList(vg_name): """Compute list of logical volumes and their size. Returns: - dictionary of all partions (key) with their size: - test1: 20.06MiB + dictionary of all partions (key) with their size (in MiB), inactive + and online status: + {'test1': ('20.06', True, True)} """ - result = utils.RunCmd(["lvs", "--noheadings", "--units=m", - "-oname,size", vg_name]) + lvs = {} + sep = '|' + result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix", + "--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) - return {} + return result.output - lvlist = [line.split() for line in result.output.splitlines()] - return dict(lvlist) + for line in result.stdout.splitlines(): + line = line.strip().rstrip(sep) + name, size, attr = line.split(sep) + if len(attr) != 6: + attr = '------' + inactive = attr[4] == '-' + online = attr[5] == 'o' + lvs[name] = (size, inactive, online) + + return lvs def ListVolumeGroups(): @@ -262,7 +306,7 @@ def NodeVolumes(): 'vg': line[3].strip(), } - return [map_line(line.split('|')) for line in result.output.splitlines()] + return [map_line(line.split('|')) for line in result.stdout.splitlines()] def BridgesExist(bridges_list): @@ -400,11 +444,10 @@ def AddOSToInstance(instance, os_disk, swap_disk): logfile) result = utils.RunCmd(command) - if result.failed: - logger.Error("os create command '%s' returned error: %s" + logger.Error("os create command '%s' returned error: %s, logfile: %s," " output: %s" % - (command, result.fail_reason, result.output)) + (command, result.fail_reason, logfile, result.output)) return False return True @@ -481,20 +524,32 @@ def _GetVGInfo(vg_name): vg_free is the free size of the volume group in MiB 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. + """ + retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"]) + retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings", "--nosuffix", "--units=m", "--separator=:", vg_name]) if retval.failed: errmsg = "volume group %s not present" % vg_name logger.Error(errmsg) - raise errors.LVMError(errmsg) - valarr = retval.stdout.strip().split(':') - retdic = { - "vg_size": int(round(float(valarr[0]), 0)), - "vg_free": int(round(float(valarr[1]), 0)), - "pv_count": int(valarr[2]), - } + return retdic + valarr = retval.stdout.strip().rstrip(':').split(':') + if len(valarr) == 3: + try: + retdic = { + "vg_size": int(round(float(valarr[0]), 0)), + "vg_free": int(round(float(valarr[1]), 0)), + "pv_count": int(valarr[2]), + } + except ValueError, err: + logger.Error("Fail to parse vgs output: %s" % str(err)) + else: + logger.Error("vgs output has the wrong number of fields (expected" + " three): %s" % str(valarr)) return retdic @@ -620,14 +675,30 @@ def RebootInstance(instance, reboot_type, extra_args): return True -def CreateBlockDevice(disk, size, on_primary, info): +def MigrateInstance(instance, target, live): + """Migrates an instance to another node. + + """ + hyper = hypervisor.GetHypervisor() + + try: + hyper.MigrateInstance(instance, target, live) + except errors.HypervisorError, err: + msg = "Failed to migrate instance: %s" % str(err) + logger.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: - bdev: a ganeti.objects.Disk object - size: the size of the physical underlying devices - do_open: if the device should be `Assemble()`-d and - `Open()`-ed after creation + 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 @@ -638,13 +709,11 @@ def CreateBlockDevice(disk, size, on_primary, info): clist = [] if disk.children: for child in disk.children: - crdev = _RecursiveAssembleBD(child, on_primary) + crdev = _RecursiveAssembleBD(child, owner, on_primary) if on_primary or disk.AssembleOnSecondary(): # we need the children open in case the device itself has to # be assembled crdev.Open() - else: - crdev.Close() clist.append(crdev) try: device = bdev.FindDevice(disk.dev_type, disk.physical_id, clist) @@ -660,10 +729,16 @@ def CreateBlockDevice(disk, size, on_primary, info): raise ValueError("Can't create child device for %s, %s" % (disk, size)) if on_primary or disk.AssembleOnSecondary(): - device.Assemble() + if not device.Assemble(): + errorstring = "Can't assemble device after creation" + logger.Error(errorstring) + raise errors.BlockDeviceError("%s, very unusual event - check the node" + " daemon logs" % errorstring) device.SetSyncSpeed(constants.SYNC_SPEED) if on_primary or disk.OpenOnSecondary(): device.Open(force=True) + DevCacheManager.UpdateCache(device.dev_path, owner, + on_primary, disk.iv_name) device.SetInfo(info) @@ -686,7 +761,10 @@ def RemoveBlockDevice(disk): logger.Info("Can't attach to device %s in remove" % disk) rdev = None if rdev is not None: + r_path = rdev.dev_path result = rdev.Remove() + if result: + DevCacheManager.RemoveCache(r_path) else: result = True if disk.children: @@ -695,7 +773,7 @@ def RemoveBlockDevice(disk): return result -def _RecursiveAssembleBD(disk, as_primary): +def _RecursiveAssembleBD(disk, owner, as_primary): """Activate a block device for an instance. This is run on the primary and secondary nodes for an instance. @@ -714,8 +792,20 @@ def _RecursiveAssembleBD(disk, as_primary): """ children = [] if disk.children: + mcn = disk.ChildrenNeeded() + if mcn == -1: + mcn = 0 # max number of Nones allowed + else: + mcn = len(disk.children) - mcn # max number of Nones for chld_disk in disk.children: - children.append(_RecursiveAssembleBD(chld_disk, as_primary)) + try: + cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary) + except errors.BlockDeviceError, err: + if children.count(None) >= mcn: + raise + cdev = None + logger.Debug("Error in child activation: %s" % str(err)) + children.append(cdev) if as_primary or disk.AssembleOnSecondary(): r_dev = bdev.AttachOrAssemble(disk.dev_type, disk.physical_id, children) @@ -723,14 +813,15 @@ def _RecursiveAssembleBD(disk, as_primary): result = r_dev if as_primary or disk.OpenOnSecondary(): r_dev.Open() - else: - r_dev.Close() + DevCacheManager.UpdateCache(r_dev.dev_path, owner, + as_primary, disk.iv_name) + else: result = True return result -def AssembleBlockDevice(disk, as_primary): +def AssembleBlockDevice(disk, owner, as_primary): """Activate a block device for an instance. This is a wrapper over _RecursiveAssembleBD. @@ -740,7 +831,7 @@ def AssembleBlockDevice(disk, as_primary): True for secondary nodes """ - result = _RecursiveAssembleBD(disk, as_primary) + result = _RecursiveAssembleBD(disk, owner, as_primary) if isinstance(result, bdev.BlockDev): result = result.dev_path return result @@ -759,7 +850,10 @@ def ShutdownBlockDevice(disk): """ r_dev = _RecursiveFindBD(disk) if r_dev is not None: + r_path = r_dev.dev_path result = r_dev.Shutdown() + if result: + DevCacheManager.RemoveCache(r_path) else: result = True if disk.children: @@ -768,35 +862,45 @@ def ShutdownBlockDevice(disk): return result -def MirrorAddChild(md_cdev, new_cdev): - """Extend an MD raid1 array. +def MirrorAddChildren(parent_cdev, new_cdevs): + """Extend a mirrored block device. """ - md_bdev = _RecursiveFindBD(md_cdev, allow_partial=True) - if md_bdev is None: - logger.Error("Can't find md device") + parent_bdev = _RecursiveFindBD(parent_cdev, allow_partial=True) + if parent_bdev is None: + logger.Error("Can't find parent device") return False - new_bdev = _RecursiveFindBD(new_cdev) - if new_bdev is None: - logger.Error("Can't find new device to add") + 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)) return False - new_bdev.Open() - md_bdev.AddChild(new_bdev) + parent_bdev.AddChildren(new_bdevs) return True -def MirrorRemoveChild(md_cdev, new_cdev): - """Reduce an MD raid1 array. +def MirrorRemoveChildren(parent_cdev, new_cdevs): + """Shrink a mirrored block device. """ - md_bdev = _RecursiveFindBD(md_cdev) - if md_bdev is None: - return False - new_bdev = _RecursiveFindBD(new_cdev) - if new_bdev is None: + parent_bdev = _RecursiveFindBD(parent_cdev) + if parent_bdev is None: + logger.Error("Can't find parent in remove children: %s" % parent_cdev) return False - new_bdev.Open() - md_bdev.RemoveChild(new_bdev.dev_path) + devs = [] + for disk in new_cdevs: + rpath = disk.StaticDevPath() + if rpath is None: + bd = _RecursiveFindBD(disk) + if bd is None: + logger.Error("Can't find dynamic device %s while removing children" % + disk) + return False + else: + devs.append(bd.dev_path) + else: + devs.append(rpath) + parent_bdev.RemoveChildren(devs) return True @@ -859,8 +963,7 @@ def FindBlockDevice(disk): rbd = _RecursiveFindBD(disk) if rbd is None: return rbd - sync_p, est_t, is_degr = rbd.GetSyncStatus() - return rbd.dev_path, rbd.major, rbd.minor, sync_p, est_t, is_degr + return (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus() def UploadFile(file_name, data, mode, uid, gid, atime, mtime): @@ -875,28 +978,20 @@ def UploadFile(file_name, data, mode, uid, gid, atime, mtime): file_name) return False - allowed_files = [constants.CLUSTER_CONF_FILE, "/etc/hosts", - constants.SSH_KNOWN_HOSTS_FILE] + 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) return False - dir_name, small_name = os.path.split(file_name) - fd, new_name = tempfile.mkstemp('.new', small_name, dir_name) - # here we need to make sure we remove the temp file, if any error - # leaves it in place - try: - os.chown(new_name, uid, gid) - os.chmod(new_name, mode) - os.write(fd, data) - os.fsync(fd) - os.utime(new_name, (atime, mtime)) - os.rename(new_name, file_name) - finally: - os.close(fd) - utils.RemoveFile(new_name) + utils.WriteFile(file_name, data=data, mode=mode, uid=uid, gid=gid, + atime=atime, mtime=mtime) return True @@ -914,27 +1009,6 @@ def _ErrnoOrStr(err): detail = str(err) 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 in search_path: - t_os_dir = os.path.sep.join([dir, name]) - if os.path.isdir(t_os_dir): - return dir - - return None def _OSOndiskVersion(name, os_dir): """Compute and return the API version of a given OS. @@ -946,7 +1020,6 @@ def _OSOndiskVersion(name, os_dir): case when this is not a valid OS name. """ - api_file = os.path.sep.join([os_dir, "ganeti_api_version"]) try: @@ -982,33 +1055,31 @@ def _OSOndiskVersion(name, os_dir): def DiagnoseOS(top_dirs=None): """Compute the validity for all OSes. - For each name in all the given top directories (if not given defaults i - to constants.OS_SEARCH_PATH it will return an object. If this is a valid - os, the object will be an instance of the object.OS class. If not, - it will be an instance of errors.InvalidOS and this signifies that - this name does not correspond to a valid OS. + 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 objects + list of OS objects """ if top_dirs is None: top_dirs = constants.OS_SEARCH_PATH result = [] - for dir in top_dirs: - if os.path.isdir(dir): + for dir_name in top_dirs: + if os.path.isdir(dir_name): try: - f_names = utils.ListVisibleFiles(dir) + f_names = utils.ListVisibleFiles(dir_name) except EnvironmentError, err: - logger.Error("Can't list the OS directory %s: %s" % (dir,str(err))) + logger.Error("Can't list the OS directory %s: %s" % + (dir_name, str(err))) break for name in f_names: try: - os_inst = OSFromDisk(name, base_dir=dir) + os_inst = OSFromDisk(name, base_dir=dir_name) result.append(os_inst) except errors.InvalidOS, err: - result.append(err) + result.append(objects.OS.FromInvalidOS(err)) return result @@ -1028,12 +1099,12 @@ def OSFromDisk(name, base_dir=None): """ 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) if api_version != constants.OS_API_VERSION: @@ -1062,7 +1133,7 @@ def OSFromDisk(name, base_dir=None): script) - return objects.OS(name=name, path=os_dir, + 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'], @@ -1070,6 +1141,32 @@ def OSFromDisk(name, base_dir=None): api_version=api_version) +def GrowBlockDevice(disk, amount): + """Grow a stack of block devices. + + This function is called recursively, with the childrens being the + first one resize. + + Args: + disk: the disk to be grown + + Returns: a tuple of (status, result), with: + status: the result (true/false) of the operation + result: the error message if the operation failed, otherwise not used + + """ + 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): """Create a snapshot copy of a block device. @@ -1093,7 +1190,7 @@ def SnapshotBlockDevice(disk): if child.size == disk.size: # return implies breaking the loop return SnapshotBlockDevice(child) - elif disk.dev_type == "lvm": + elif disk.dev_type == constants.LD_LV: r_dev = _RecursiveFindBD(disk) if r_dev is not None: # let's stay on the safe side and ask for the full size, for now @@ -1102,7 +1199,7 @@ def SnapshotBlockDevice(disk): return None else: raise errors.ProgrammerError("Cannot snapshot non-lvm block device" - "'%s' of type '%s'" % + " '%s' of type '%s'" % (disk.unique_id, disk.dev_type)) @@ -1147,9 +1244,8 @@ def ExportSnapshot(disk, dest_node, instance): destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s", destdir, destdir, destfile) - remotecmd = ssh.BuildSSHCmd(dest_node, constants.GANETI_RUNAS, destcmd) - - + remotecmd = _GetSshRunner().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)]) @@ -1193,13 +1289,17 @@ def FinalizeExport(instance, snap_disks): config.set(constants.INISECT_INS, 'memory', '%d' % instance.memory) config.set(constants.INISECT_INS, 'vcpus', '%d' % instance.vcpus) config.set(constants.INISECT_INS, 'disk_template', instance.disk_template) + + nic_count = 0 for nic_count, nic in enumerate(instance.nics): config.set(constants.INISECT_INS, 'nic%d_mac' % nic_count, '%s' % nic.mac) config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip) + config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count, '%s' % nic.bridge) # TODO: redundant: on load can read nics until it doesn't exist config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_count) + disk_count = 0 for disk_count, disk in enumerate(snap_disks): config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count, ('%s' % disk.iv_name)) @@ -1289,7 +1389,8 @@ def ImportOSIntoInstance(instance, os_disk, swap_disk, src_node, src_image): os.mkdir(constants.LOG_OS_DIR, 0750) destcmd = utils.BuildShellCmd('cat %s', src_image) - remotecmd = ssh.BuildSSHCmd(src_node, constants.GANETI_RUNAS, destcmd) + remotecmd = _GetSshRunner().BuildCmd(src_node, constants.GANETI_RUNAS, + destcmd) comprcmd = "gunzip" impcmd = utils.BuildShellCmd("(cd %s; %s -i %s -b %s -s %s &>%s)", @@ -1339,6 +1440,187 @@ def RemoveExport(export): return True +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). + + """ + result = True + for disk, unique_id in devlist: + dev = _RecursiveFindBD(disk) + if dev is None: + result = False + continue + try: + old_rpath = dev.dev_path + dev.Rename(unique_id) + new_rpath = dev.dev_path + if old_rpath != new_rpath: + DevCacheManager.RemoveCache(old_rpath) + # FIXME: we should add the new cache information here, like: + # DevCacheManager.UpdateCache(new_rpath, owner, ...) + # but we don't have the owner here - maybe parse from existing + # 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)) + result = False + return result + + +def _TransformFileStorageDir(file_storage_dir): + """Checks whether given file_storage_dir is valid. + + Checks wheter the given file_storage_dir is within the cluster-wide + 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 + + """ + file_storage_dir = os.path.normpath(file_storage_dir) + base_file_storage_dir = ssconf.SimpleStore().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)) + return None + return file_storage_dir + + +def CreateFileStorageDir(file_storage_dir): + """Create file storage directory. + + Args: + file_storage_dir: string containing the path + + Returns: + tuple with first element a boolean indicating wheter dir + creation was successful or not + + """ + file_storage_dir = _TransformFileStorageDir(file_storage_dir) + result = True, + if not file_storage_dir: + result = False, + 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) + 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)) + result = False, + return result + + +def RemoveFileStorageDir(file_storage_dir): + """Remove file storage directory. + + 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 + + """ + file_storage_dir = _TransformFileStorageDir(file_storage_dir) + result = True, + if not file_storage_dir: + result = False, + 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) + 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)) + 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 + + """ + old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir) + new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir) + result = True, + if not old_file_storage_dir or not new_file_storage_dir: + result = False, + else: + if not os.path.exists(new_file_storage_dir): + if os.path.isdir(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)) + result = False, + else: + logger.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) + result = False, + return result + + +def CloseBlockDevices(disks): + """Closes the given block devices. + + This means they will be switched to secondary mode (in case of DRBD). + + """ + 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") + + class HooksRunner(object): """Hook runner. @@ -1354,9 +1636,6 @@ class HooksRunner(object): 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 """ if hooks_base_dir is None: @@ -1368,7 +1647,6 @@ class HooksRunner(object): """Exec one hook script. Args: - - phase: the phase - script: the full path to the script - env: the environment with which to exec the script @@ -1379,7 +1657,7 @@ class HooksRunner(object): 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) + shell=False, cwd="/", env=env) output = "" try: output = child.stdout.read(4096) @@ -1448,3 +1726,98 @@ class HooksRunner(object): rr.append(("%s/%s" % (subdir, relname), rrval, output)) 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. + + Return value: tuple of: + - run status (one of the IARUN_ constants) + - stdout + - stderr + - fail reason (as from 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. + + """ + _DEV_PREFIX = "/dev/" + _ROOT_DIR = constants.BDEV_CACHE_DIR + + @classmethod + def _ConvertPath(cls, dev_path): + """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 + + """ + 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) + return fpath + + @classmethod + def UpdateCache(cls, dev_path, owner, on_primary, iv_name): + """Updates the cache information for a given device. + + """ + if dev_path is None: + logger.Error("DevCacheManager.UpdateCache got a None dev_path") + return + fpath = cls._ConvertPath(dev_path) + if on_primary: + state = "primary" + else: + state = "secondary" + if iv_name is None: + iv_name = "not_visible" + fdata = "%s %s %s\n" % (str(owner), state, iv_name) + try: + utils.WriteFile(fpath, data=fdata) + except EnvironmentError, err: + logger.Error("Can't update bdev cache for %s, error %s" % + (dev_path, str(err))) + + @classmethod + def RemoveCache(cls, dev_path): + """Remove data for a dev_path. + + """ + if dev_path is None: + logger.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)))