X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/ff73280e9c5c3d8d7930949fadca997cbccddd22..8c229cc7e24f1c7075b38ff0e21aab7c75299c4d:/lib/backend.py diff --git a/lib/backend.py b/lib/backend.py index f54c041..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"]) @@ -137,6 +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. @@ -199,10 +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 @@ -210,19 +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.stdout.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(): @@ -479,20 +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 @@ -618,14 +675,30 @@ def RebootInstance(instance, reboot_type, extra_args): return True +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 @@ -641,8 +714,6 @@ def CreateBlockDevice(disk, size, owner, on_primary, info): # 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) @@ -742,8 +813,6 @@ def _RecursiveAssembleBD(disk, owner, 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) @@ -909,28 +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 @@ -949,28 +1010,6 @@ def _ErrnoOrStr(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_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. @@ -1060,12 +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: @@ -1102,6 +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. @@ -1179,9 +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)]) @@ -1225,13 +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)) @@ -1321,7 +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)", @@ -1403,6 +1472,155 @@ def RenameBlockDevices(devlist): 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. @@ -1418,9 +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: @@ -1432,7 +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 @@ -1443,7 +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) @@ -1514,8 +1728,44 @@ class HooksRunner(object): 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 chache of block device information. + """Simple class for managing a cache of block device information. """ _DEV_PREFIX = "/dev/"