X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/ea9c753d1c5e376f14b30198f68124369cc2a09f..22001b7894bd406e2180bd6aafc0ca987c40cb6b:/lib/backend.py diff --git a/lib/backend.py b/lib/backend.py index 5b77dae..2c5e5ae 100644 --- a/lib/backend.py +++ b/lib/backend.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -60,14 +60,19 @@ from ganeti import ssconf from ganeti import serializer from ganeti import netutils from ganeti import runtime +from ganeti import compat +from ganeti import pathutils +from ganeti import vcluster +from ganeti import ht +from ganeti import hooksmaster _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id" -_ALLOWED_CLEAN_DIRS = frozenset([ - constants.DATA_DIR, - constants.JOB_QUEUE_ARCHIVE_DIR, - constants.QUEUE_DIR, - constants.CRYPTO_KEYS_DIR, +_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([ + pathutils.DATA_DIR, + pathutils.JOB_QUEUE_ARCHIVE_DIR, + pathutils.QUEUE_DIR, + pathutils.CRYPTO_KEYS_DIR, ]) _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60 _X509_KEY_FILE = "key" @@ -77,7 +82,24 @@ _IES_PID_FILE = "pid" _IES_CA_FILE = "ca" #: Valid LVS output line regex -_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6})\|?$") +_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$") + +# Actions for the master setup script +_MASTER_START = "start" +_MASTER_STOP = "stop" + +#: Maximum file permissions for restricted command directory and executables +_RCMD_MAX_MODE = (stat.S_IRWXU | + stat.S_IRGRP | stat.S_IXGRP | + stat.S_IROTH | stat.S_IXOTH) + +#: Delay before returning an error for restricted commands +_RCMD_INVALID_DELAY = 10 + +#: How long to wait to acquire lock for restricted commands (shorter than +#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many +#: command requests arrive +_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8 class RPCFail(Exception): @@ -88,6 +110,34 @@ class RPCFail(Exception): """ +def _GetInstReasonFilename(instance_name): + """Path of the file containing the reason of the instance status change. + + @type instance_name: string + @param instance_name: The name of the instance + @rtype: string + @return: The path of the file + + """ + return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name) + + +def _StoreInstReasonTrail(instance_name, trail): + """Serialize a reason trail related to an instance change of state to file. + + The exact location of the file depends on the name of the instance and on + the configuration of the Ganeti cluster defined at deploy time. + + @type instance_name: string + @param instance_name: The name of the instance + @rtype: None + + """ + json = serializer.DumpJson(trail) + filename = _GetInstReasonFilename(instance_name) + utils.WriteFile(filename, data=json) + + def _Fail(msg, *args, **kwargs): """Log an error and the raise an RPCFail exception. @@ -191,21 +241,24 @@ def _BuildUploadFileList(): """ allowed_files = set([ - constants.CLUSTER_CONF_FILE, - constants.ETC_HOSTS, - constants.SSH_KNOWN_HOSTS_FILE, - constants.VNC_PASSWORD_FILE, - constants.RAPI_CERT_FILE, - constants.SPICE_CERT_FILE, - constants.SPICE_CACERT_FILE, - constants.RAPI_USERS_FILE, - constants.CONFD_HMAC_KEY, - constants.CLUSTER_DOMAIN_SECRET_FILE, + pathutils.CLUSTER_CONF_FILE, + pathutils.ETC_HOSTS, + pathutils.SSH_KNOWN_HOSTS_FILE, + pathutils.VNC_PASSWORD_FILE, + pathutils.RAPI_CERT_FILE, + pathutils.SPICE_CERT_FILE, + pathutils.SPICE_CACERT_FILE, + pathutils.RAPI_USERS_FILE, + pathutils.CONFD_HMAC_KEY, + pathutils.CLUSTER_DOMAIN_SECRET_FILE, ]) for hv_name in constants.HYPER_TYPES: hv_class = hypervisor.GetHypervisorClass(hv_name) - allowed_files.update(hv_class.GetAncillaryFiles()) + allowed_files.update(hv_class.GetAncillaryFiles()[0]) + + assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \ + "Allowed file storage paths should never be uploaded via RPC" return frozenset(allowed_files) @@ -220,8 +273,8 @@ def JobQueuePurge(): @return: True, None """ - _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE]) - _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR) + _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE]) + _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR) def GetMasterInfo(): @@ -231,7 +284,8 @@ def GetMasterInfo(): for consumption here or from the node daemon. @rtype: tuple - @return: master_netdev, master_ip, master_name, primary_ip_family + @return: master_netdev, master_ip, master_name, primary_ip_family, + master_netmask @raise RPCFail: in case of errors """ @@ -239,54 +293,118 @@ def GetMasterInfo(): cfg = _GetConfig() master_netdev = cfg.GetMasterNetdev() master_ip = cfg.GetMasterIP() + master_netmask = cfg.GetMasterNetmask() master_node = cfg.GetMasterNode() primary_ip_family = cfg.GetPrimaryIPFamily() except errors.ConfigurationError, err: _Fail("Cluster configuration incomplete: %s", err, exc=True) - return (master_netdev, master_ip, master_node, primary_ip_family) + return (master_netdev, master_ip, master_node, primary_ip_family, + master_netmask) -def ActivateMasterIp(): - """Activate the IP address of the master daemon. +def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn): + """Decorator that runs hooks before and after the decorated function. + + @type hook_opcode: string + @param hook_opcode: opcode of the hook + @type hooks_path: string + @param hooks_path: path of the hooks + @type env_builder_fn: function + @param env_builder_fn: function that returns a dictionary containing the + environment variables for the hooks. Will get all the parameters of the + decorated function. + @raise RPCFail: in case of pre-hook failure """ - # GetMasterInfo will raise an exception if not able to return data - master_netdev, master_ip, _, family = GetMasterInfo() + def decorator(fn): + def wrapper(*args, **kwargs): + _, myself = ssconf.GetMasterAndMyself() + nodes = ([myself], [myself]) # these hooks run locally + + env_fn = compat.partial(env_builder_fn, *args, **kwargs) + + cfg = _GetConfig() + hr = HooksRunner() + hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes, + hr.RunLocalHooks, None, env_fn, + logging.warning, cfg.GetClusterName(), + cfg.GetMasterNode()) + hm.RunPhase(constants.HOOKS_PHASE_PRE) + result = fn(*args, **kwargs) + hm.RunPhase(constants.HOOKS_PHASE_POST) + + return result + return wrapper + return decorator + + +def _BuildMasterIpEnv(master_params, use_external_mip_script=None): + """Builds environment variables for master IP hooks. + + @type master_params: L{objects.MasterNetworkParameters} + @param master_params: network parameters of the master + @type use_external_mip_script: boolean + @param use_external_mip_script: whether to use an external master IP + address setup script (unused, but necessary per the implementation of the + _RunLocalHooks decorator) - err_msg = None - if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT): - if netutils.IPAddress.Own(master_ip): - # we already have the ip: - logging.debug("Master IP already configured, doing nothing") - else: - err_msg = "Someone else has the master ip, not activating" - logging.error(err_msg) + """ + # pylint: disable=W0613 + ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family) + env = { + "MASTER_NETDEV": master_params.netdev, + "MASTER_IP": master_params.ip, + "MASTER_NETMASK": str(master_params.netmask), + "CLUSTER_IP_VERSION": str(ver), + } + + return env + + +def _RunMasterSetupScript(master_params, action, use_external_mip_script): + """Execute the master IP address setup script. + + @type master_params: L{objects.MasterNetworkParameters} + @param master_params: network parameters of the master + @type action: string + @param action: action to pass to the script. Must be one of + L{backend._MASTER_START} or L{backend._MASTER_STOP} + @type use_external_mip_script: boolean + @param use_external_mip_script: whether to use an external master IP + address setup script + @raise backend.RPCFail: if there are errors during the execution of the + script + + """ + env = _BuildMasterIpEnv(master_params) + + if use_external_mip_script: + setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT else: - ipcls = netutils.IP4Address - if family == netutils.IP6Address.family: - ipcls = netutils.IP6Address - - result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add", - "%s/%d" % (master_ip, ipcls.iplen), - "dev", master_netdev, "label", - "%s:0" % master_netdev]) - if result.failed: - err_msg = "Can't activate master IP: %s" % result.output - logging.error(err_msg) - - # we ignore the exit code of the following cmds - if ipcls == netutils.IP4Address: - utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, "-s", - master_ip, master_ip]) - elif ipcls == netutils.IP6Address: - try: - utils.RunCmd(["ndisc6", "-q", "-r 3", master_ip, master_netdev]) - except errors.OpExecError: - # TODO: Better error reporting - logging.warning("Can't execute ndisc6, please install if missing") + setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT - if err_msg: - _Fail(err_msg) + result = utils.RunCmd([setup_script, action], env=env, reset_env=True) + + if result.failed: + _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" % + (action, result.exit_code, result.output), log=True) + + +@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup", + _BuildMasterIpEnv) +def ActivateMasterIp(master_params, use_external_mip_script): + """Activate the IP address of the master daemon. + + @type master_params: L{objects.MasterNetworkParameters} + @param master_params: network parameters of the master + @type use_external_mip_script: boolean + @param use_external_mip_script: whether to use an external master IP + address setup script + @raise RPCFail: in case of errors during the IP startup + + """ + _RunMasterSetupScript(master_params, _MASTER_START, + use_external_mip_script) def StartMasterDaemons(no_voting): @@ -310,33 +428,28 @@ def StartMasterDaemons(no_voting): "EXTRA_MASTERD_ARGS": masterd_args, } - result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env) + result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env) if result.failed: msg = "Can't start Ganeti master: %s" % result.output logging.error(msg) _Fail(msg) -def DeactivateMasterIp(): +@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown", + _BuildMasterIpEnv) +def DeactivateMasterIp(master_params, use_external_mip_script): """Deactivate the master IP on this node. - """ - # TODO: log and report back to the caller the error failures; we - # need to decide in which case we fail the RPC for this + @type master_params: L{objects.MasterNetworkParameters} + @param master_params: network parameters of the master + @type use_external_mip_script: boolean + @param use_external_mip_script: whether to use an external master IP + address setup script + @raise RPCFail: in case of errors during the IP turndown - # GetMasterInfo will raise an exception if not able to return data - master_netdev, master_ip, _, family = GetMasterInfo() - - ipcls = netutils.IP4Address - if family == netutils.IP6Address.family: - ipcls = netutils.IP6Address - - result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del", - "%s/%d" % (master_ip, ipcls.iplen), - "dev", master_netdev]) - if result.failed: - logging.error("Can't remove the master IP, error: %s", result.output) - # but otherwise ignore the failure + """ + _RunMasterSetupScript(master_params, _MASTER_STOP, + use_external_mip_script) def StopMasterDaemons(): @@ -350,13 +463,44 @@ def StopMasterDaemons(): # TODO: log and report back to the caller the error failures; we # need to decide in which case we fail the RPC for this - result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"]) + result = utils.RunCmd([pathutils.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 ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev): + """Change the netmask of the master IP. + + @param old_netmask: the old value of the netmask + @param netmask: the new value of the netmask + @param master_ip: the master IP + @param master_netdev: the master network device + + """ + if old_netmask == netmask: + return + + if not netutils.IPAddress.Own(master_ip): + _Fail("The master IP address is not up, not attempting to change its" + " netmask") + + result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add", + "%s/%s" % (master_ip, netmask), + "dev", master_netdev, "label", + "%s:0" % master_netdev]) + if result.failed: + _Fail("Could not set the new netmask on the master IP address") + + result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del", + "%s/%s" % (master_ip, old_netmask), + "dev", master_netdev, "label", + "%s:0" % master_netdev]) + if result.failed: + _Fail("Could not bring down the master IP address with the old netmask") + + def EtcHostsModify(mode, host, ip): """Modify a host entry in /etc/hosts. @@ -392,13 +536,13 @@ def LeaveCluster(modify_ssh_setup): @param modify_ssh_setup: boolean """ - _CleanDirectory(constants.DATA_DIR) - _CleanDirectory(constants.CRYPTO_KEYS_DIR) + _CleanDirectory(pathutils.DATA_DIR) + _CleanDirectory(pathutils.CRYPTO_KEYS_DIR) JobQueuePurge() if modify_ssh_setup: try: - priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) + priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER) utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) @@ -408,15 +552,15 @@ def LeaveCluster(modify_ssh_setup): logging.exception("Error while processing ssh files") try: - utils.RemoveFile(constants.CONFD_HMAC_KEY) - utils.RemoveFile(constants.RAPI_CERT_FILE) - utils.RemoveFile(constants.SPICE_CERT_FILE) - utils.RemoveFile(constants.SPICE_CACERT_FILE) - utils.RemoveFile(constants.NODED_CERT_FILE) + utils.RemoveFile(pathutils.CONFD_HMAC_KEY) + utils.RemoveFile(pathutils.RAPI_CERT_FILE) + utils.RemoveFile(pathutils.SPICE_CERT_FILE) + utils.RemoveFile(pathutils.SPICE_CACERT_FILE) + utils.RemoveFile(pathutils.NODED_CERT_FILE) except: # pylint: disable=W0702 logging.exception("Error while removing cluster secrets") - result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD]) + result = utils.RunCmd([pathutils.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) @@ -425,44 +569,90 @@ def LeaveCluster(modify_ssh_setup): raise errors.QuitGanetiException(True, "Shutdown scheduled") -def GetNodeInfo(vgname, hypervisor_type): +def _GetVgInfo(name, excl_stor): + """Retrieves information about a LVM volume group. + + """ + # TODO: GetVGInfo supports returning information for multiple VGs at once + vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor) + if vginfo: + vg_free = int(round(vginfo[0][0], 0)) + vg_size = int(round(vginfo[0][1], 0)) + else: + vg_free = None + vg_size = None + + return { + "name": name, + "vg_free": vg_free, + "vg_size": vg_size, + } + + +def _GetHvInfo(name): + """Retrieves node information from a hypervisor. + + The information returned depends on the hypervisor. Common items: + + - vg_size is the size of the configured volume group in MiB + - vg_free is the free size of the volume group in MiB + - memory_dom0 is the memory allocated for domain0 in MiB + - memory_free is the currently available (free) ram in MiB + - memory_total is the total number of ram in MiB + - hv_version: the hypervisor version, if available + + """ + return hypervisor.GetHypervisor(name).GetNodeInfo() + + +def _GetNamedNodeInfo(names, fn): + """Calls C{fn} for all names in C{names} and returns a dictionary. + + @rtype: None or dict + + """ + if names is None: + return None + else: + return map(fn, names) + + +def GetNodeInfo(vg_names, hv_names, excl_stor): """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 - @type hypervisor_type: C{str} - @param hypervisor_type: the name of the hypervisor to ask for - memory information - @rtype: C{dict} - @return: dictionary with the following keys: - - vg_size is the size of the configured volume group in MiB - - vg_free is the free size of the volume group in MiB - - memory_dom0 is the memory allocated for domain0 in MiB - - memory_free is the currently available (free) ram in MiB - - memory_total is the total number of ram in MiB - - hv_version: the hypervisor version, if available + @type vg_names: list of string + @param vg_names: Names of the volume groups to ask for disk space information + @type hv_names: list of string + @param hv_names: Names of the hypervisors to ask for node information + @type excl_stor: boolean + @param excl_stor: Whether exclusive_storage is active + @rtype: tuple; (string, None/dict, None/dict) + @return: Tuple containing boot ID, volume group information and hypervisor + information """ - outputarray = {} + bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n") + vg_info = _GetNamedNodeInfo(vg_names, (lambda vg: _GetVgInfo(vg, excl_stor))) + hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo) + + return (bootid, vg_info, hv_info) - if vgname is not None: - vginfo = bdev.LogicalVolume.GetVGInfo([vgname]) - vg_free = vg_size = None - if vginfo: - vg_free = int(round(vginfo[0][0], 0)) - vg_size = int(round(vginfo[0][1], 0)) - outputarray["vg_size"] = vg_size - outputarray["vg_free"] = vg_free - if hypervisor_type is not None: - hyper = hypervisor.GetHypervisor(hypervisor_type) - hyp_info = hyper.GetNodeInfo() - if hyp_info is not None: - outputarray.update(hyp_info) +def _CheckExclusivePvs(pvi_list): + """Check that PVs are not shared among LVs - outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n") + @type pvi_list: list of L{objects.LvmPvInfo} objects + @param pvi_list: information about the PVs - return outputarray + @rtype: list of tuples (string, list of strings) + @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...]) + + """ + res = [] + for pvi in pvi_list: + if len(pvi.lv_list) > 1: + res.append((pvi.name, pvi.lv_list)) + return res def VerifyNode(what, cluster_name): @@ -518,8 +708,11 @@ def VerifyNode(what, cluster_name): tmp.append((source, hv_name, str(err))) if constants.NV_FILELIST in what: - result[constants.NV_FILELIST] = utils.FingerprintFiles( - what[constants.NV_FILELIST]) + fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath, + what[constants.NV_FILELIST])) + result[constants.NV_FILELIST] = \ + dict((vcluster.MakeVirtualPath(key), value) + for (key, value) in fingerprints.items()) if constants.NV_NODELIST in what: (nodes, bynode) = what[constants.NV_NODELIST] @@ -574,7 +767,12 @@ def VerifyNode(what, cluster_name): else: source = None result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port, - source=source) + source=source) + + if constants.NV_USERSCRIPTS in what: + result[constants.NV_USERSCRIPTS] = \ + [script for script in what[constants.NV_USERSCRIPTS] + if not utils.IsExecutable(script)] if constants.NV_OOB_PATHS in what: result[constants.NV_OOB_PATHS] = tmp = [] @@ -611,9 +809,16 @@ def VerifyNode(what, cluster_name): result[constants.NV_VGLIST] = utils.ListVolumeGroups() if constants.NV_PVLIST in what and vm_capable: - result[constants.NV_PVLIST] = \ - bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST], - filter_allocatable=False) + check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what + val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST], + filter_allocatable=False, + include_lvs=check_exclusive_pvs) + if check_exclusive_pvs: + result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val) + for pvi in val: + # Avoid sending useless data on the wire + pvi.lv_list = [] + result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val) if constants.NV_VERSION in what: result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION, @@ -663,6 +868,11 @@ def VerifyNode(what, cluster_name): result[constants.NV_BRIDGES] = [bridge for bridge in what[constants.NV_BRIDGES] if not utils.BridgeExists(bridge)] + + if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name: + result[constants.NV_FILE_STORAGE_PATHS] = \ + bdev.ComputeWrongFileStoragePaths() + return result @@ -860,6 +1070,7 @@ def GetInstanceInfo(instance, hname): - memory: memory size of instance (int) - state: xen state of instance (string) - time: cpu time of instance (float) + - vcpus: the number of vcpus (int) """ output = {} @@ -867,6 +1078,7 @@ def GetInstanceInfo(instance, hname): iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance) if iinfo is not None: output["memory"] = iinfo[2] + output["vcpus"] = iinfo[3] output["state"] = iinfo[4] output["time"] = iinfo[5] @@ -965,7 +1177,7 @@ def _InstanceLogName(kind, os_name, instance, component): c_msg = "" base = ("%s-%s-%s%s-%s.log" % (kind, os_name, instance, c_msg, utils.TimestampForFilename())) - return utils.PathJoin(constants.LOG_OS_DIR, base) + return utils.PathJoin(pathutils.LOG_OS_DIR, base) def InstanceOsAdd(instance, reinstall, debug): @@ -1033,9 +1245,16 @@ def RunRenameInstance(instance, old_name, debug): " log file:\n%s", result.fail_reason, "\n".join(lines), log=False) -def _GetBlockDevSymlinkPath(instance_name, idx): - return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" % - (instance_name, constants.DISK_SEPARATOR, idx)) +def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None): + """Returns symlink path for block device. + + """ + if _dir is None: + _dir = pathutils.DISK_LINKS_DIR + + return utils.PathJoin(_dir, + ("%s%s%s" % + (instance_name, constants.DISK_SEPARATOR, idx))) def _SymlinkBlockDev(instance_name, device_path, idx): @@ -1109,13 +1328,17 @@ def _GatherAndLinkBlockDevs(instance): return block_devices -def StartInstance(instance, startup_paused): +def StartInstance(instance, startup_paused, reason, store_reason=True): """Start an instance. @type instance: L{objects.Instance} @param instance: the instance object @type startup_paused: bool @param instance: pause instance at startup? + @type reason: list of reasons + @param reason: the reason trail for this startup + @type store_reason: boolean + @param store_reason: whether to store the shutdown reason trail on file @rtype: None """ @@ -1129,6 +1352,8 @@ def StartInstance(instance, startup_paused): block_devices = _GatherAndLinkBlockDevs(instance) hyper = hypervisor.GetHypervisor(instance.hypervisor) hyper.StartInstance(instance, block_devices, startup_paused) + if store_reason: + _StoreInstReasonTrail(instance.name, reason) except errors.BlockDeviceError, err: _Fail("Block device error: %s", err, exc=True) except errors.HypervisorError, err: @@ -1136,7 +1361,7 @@ def StartInstance(instance, startup_paused): _Fail("Hypervisor error: %s", err, exc=True) -def InstanceShutdown(instance, timeout): +def InstanceShutdown(instance, timeout, reason, store_reason=True): """Shut an instance down. @note: this functions uses polling with a hardcoded timeout. @@ -1145,6 +1370,10 @@ def InstanceShutdown(instance, timeout): @param instance: the instance object @type timeout: integer @param timeout: maximum timeout for soft shutdown + @type reason: list of reasons + @param reason: the reason trail for this shutdown + @type store_reason: boolean + @param store_reason: whether to store the shutdown reason trail on file @rtype: None """ @@ -1166,6 +1395,8 @@ def InstanceShutdown(instance, timeout): try: hyper.StopInstance(instance, retry=self.tried_once) + if store_reason: + _StoreInstReasonTrail(instance.name, reason) except errors.HypervisorError, err: if iname not in hyper.ListInstances(): # if the instance is no longer existing, consider this a @@ -1205,7 +1436,7 @@ def InstanceShutdown(instance, timeout): _RemoveBlockDevLinks(iname, instance.disks) -def InstanceReboot(instance, reboot_type, shutdown_timeout): +def InstanceReboot(instance, reboot_type, shutdown_timeout, reason): """Reboot an instance. @type instance: L{objects.Instance} @@ -1223,6 +1454,8 @@ def InstanceReboot(instance, reboot_type, shutdown_timeout): instance (instead of a call_instance_reboot RPC) @type shutdown_timeout: integer @param shutdown_timeout: maximum timeout for soft shutdown + @type reason: list of reasons + @param reason: the reason trail for this reboot @rtype: None """ @@ -1239,14 +1472,37 @@ def InstanceReboot(instance, reboot_type, shutdown_timeout): _Fail("Failed to soft reboot instance %s: %s", instance.name, err) elif reboot_type == constants.INSTANCE_REBOOT_HARD: try: - InstanceShutdown(instance, shutdown_timeout) - return StartInstance(instance, False) + InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False) + result = StartInstance(instance, False, reason, store_reason=False) + _StoreInstReasonTrail(instance.name, reason) + return result except errors.HypervisorError, err: _Fail("Failed to hard reboot instance %s: %s", instance.name, err) else: _Fail("Invalid reboot_type received: %s", reboot_type) +def InstanceBalloonMemory(instance, memory): + """Resize an instance's memory. + + @type instance: L{objects.Instance} + @param instance: the instance object + @type memory: int + @param memory: new memory amount in MB + @rtype: None + + """ + hyper = hypervisor.GetHypervisor(instance.hypervisor) + running = hyper.ListInstances() + if instance.name not in running: + logging.info("Instance %s is not running, cannot balloon", instance.name) + return + try: + hyper.BalloonInstanceMemory(instance, memory) + except errors.HypervisorError, err: + _Fail("Failed to balloon instance memory: %s", err, exc=True) + + def MigrationInfo(instance): """Gather information about an instance to be migrated. @@ -1370,7 +1626,55 @@ def GetMigrationStatus(instance): _Fail("Failed to get migration status: %s", err, exc=True) -def BlockdevCreate(disk, size, owner, on_primary, info): +def HotplugDevice(instance, action, dev_type, device, extra, seq): + """Hotplug a device + + Hotplug is currently supported only for KVM Hypervisor. + @type instance: L{objects.Instance} + @param instance: the instance to which we hotplug a device + @type action: string + @param action: the hotplug action to perform + @type dev_type: string + @param dev_type: the device type to hotplug + @type device: either L{objects.NIC} or L{objects.Disk} + @param device: the device object to hotplug + @type extra: string + @param extra: extra info used by hotplug code (e.g. disk link) + @type seq: int + @param seq: the index of the device from master perspective + @raise RPCFail: in case instance does not have KVM hypervisor + + """ + hyper = hypervisor.GetHypervisor(instance.hypervisor) + try: + hyper.VerifyHotplugSupport(instance, action, dev_type) + except errors.HotplugError, err: + _Fail("Hotplug is not supported: %s", err) + + if action == constants.HOTPLUG_ACTION_ADD: + fn = hyper.HotAddDevice + elif action == constants.HOTPLUG_ACTION_REMOVE: + fn = hyper.HotDelDevice + elif action == constants.HOTPLUG_ACTION_MODIFY: + fn = hyper.HotModDevice + else: + assert action in constants.HOTPLUG_ALL_ACTIONS + + return fn(instance, dev_type, device, extra, seq) + + +def HotplugSupported(instance): + """Checks if hotplug is generally supported. + + """ + hyper = hypervisor.GetHypervisor(instance.hypervisor) + try: + hyper.HotplugSupported(instance) + except errors.HotplugError, err: + _Fail("Hotplug is not supported: %s", err) + + +def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor): """Creates a block device for an instance. @type disk: L{objects.Disk} @@ -1385,6 +1689,8 @@ def BlockdevCreate(disk, size, owner, on_primary, info): @type info: string @param info: string that will be sent to the physical device creation, used for example to set (LVM) tags on LVs + @type excl_stor: boolean + @param excl_stor: Whether exclusive_storage is active @return: the new unique_id of the device (this can sometime be computed only after creation), or None. On secondary nodes, @@ -1411,7 +1717,7 @@ def BlockdevCreate(disk, size, owner, on_primary, info): clist.append(crdev) try: - device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size) + device = bdev.Create(disk, clist, excl_stor) except errors.BlockDeviceError, err: _Fail("Can't create block device: %s", err) @@ -1420,7 +1726,6 @@ def BlockdevCreate(disk, size, owner, on_primary, info): device.Assemble() except errors.BlockDeviceError, err: _Fail("Can't assemble device after creation, unusual event: %s", err) - device.SetSyncSpeed(constants.SYNC_SPEED) if on_primary or disk.OpenOnSecondary(): try: device.Open(force=True) @@ -1442,8 +1747,13 @@ def _WipeDevice(path, offset, size): @param size: The size in MiB to write """ + # Internal sizes are always in Mebibytes; if the following "dd" command + # should use a different block size the offset and size given to this + # function must be adjusted accordingly before being passed to "dd". + block_size = 1024 * 1024 + cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset, - "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path, + "bs=%s" % block_size, "oflag=direct", "of=%s" % path, "count=%d" % size] result = utils.RunCmd(cmd) @@ -1472,6 +1782,10 @@ def BlockdevWipe(disk, offset, size): _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name) # Do cross verify some of the parameters + if offset < 0: + _Fail("Negative offset") + if size < 0: + _Fail("Negative size") if offset > rdev.size: _Fail("Offset is bigger than device size") if (offset + size) > rdev.size: @@ -1535,10 +1849,18 @@ def BlockdevRemove(disk): rdev = None if rdev is not None: r_path = rdev.dev_path - try: - rdev.Remove() - except errors.BlockDeviceError, err: - msgs.append(str(err)) + + def _TryRemove(): + try: + rdev.Remove() + return [] + except errors.BlockDeviceError, err: + return [str(err)] + + msgs.extend(utils.SimpleRetry([], _TryRemove, + constants.DISK_REMOVE_RETRY_INTERVAL, + constants.DISK_REMOVE_RETRY_TIMEOUT)) + if not msgs: DevCacheManager.RemoveCache(r_path) @@ -1594,8 +1916,7 @@ def _RecursiveAssembleBD(disk, owner, as_primary): children.append(cdev) if as_primary or disk.AssembleOnSecondary(): - r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size) - r_dev.SetSyncSpeed(constants.SYNC_SPEED) + r_dev = bdev.Assemble(disk, children) result = r_dev if as_primary or disk.OpenOnSecondary(): r_dev.Open() @@ -1613,23 +1934,28 @@ def BlockdevAssemble(disk, owner, as_primary, idx): This is a wrapper over _RecursiveAssembleBD. @rtype: str or boolean - @return: a C{/dev/...} path for primary nodes, and - C{True} for secondary nodes + @return: a tuple with the C{/dev/...} path and the created symlink + for primary nodes, and (C{True}, C{True}) for secondary nodes """ try: result = _RecursiveAssembleBD(disk, owner, as_primary) if isinstance(result, bdev.BlockDev): # pylint: disable=E1103 - result = result.dev_path + dev_path = result.dev_path + link_name = None if as_primary: - _SymlinkBlockDev(owner, result, idx) + link_name = _SymlinkBlockDev(owner, dev_path, idx) + elif result: + return result, result + else: + _Fail("Unexpected result from _RecursiveAssembleBD") except errors.BlockDeviceError, err: _Fail("Error while assembling disk: %s", err, exc=True) except OSError, err: _Fail("Error while symlinking disk: %s", err, exc=True) - return result + return dev_path, link_name def BlockdevShutdown(disk): @@ -1788,7 +2114,7 @@ def _RecursiveFindBD(disk): for chdisk in disk.children: children.append(_RecursiveFindBD(chdisk)) - return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size) + return bdev.FindDevice(disk, children) def _OpenRealBD(disk): @@ -1887,7 +2213,7 @@ def BlockdevExport(disk, dest_node, dest_path, cluster_name): " oflag=dsync", dest_path) remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node, - constants.GANETI_RUNAS, + constants.SSH_LOGIN_USER, destcmd) # all commands have been checked, so we're safe to combine them @@ -1923,6 +2249,8 @@ def UploadFile(file_name, data, mode, uid, gid, atime, mtime): @rtype: None """ + file_name = vcluster.LocalizeVirtualPath(file_name) + if not os.path.isabs(file_name): _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name) @@ -1965,33 +2293,6 @@ def RunOob(oob_program, command, node, timeout): return result.stdout -def WriteSsconfFiles(values): - """Update all ssconf files. - - Wrapper around the SimpleStore.WriteFiles. - - """ - ssconf.SimpleStore().WriteFiles(values) - - -def _ErrnoOrStr(err): - """Format an EnvironmentError exception. - - If the L{err} argument has an errno attribute, it will be looked up - and converted into a textual C{E...} description. Otherwise the - string representation of the error will be returned. - - @type err: L{EnvironmentError} - @param err: the exception to format - - """ - if hasattr(err, "errno"): - detail = errno.errorcode[err.errno] - else: - detail = str(err) - return detail - - def _OSOndiskAPIVersion(os_dir): """Compute and return the API version of a given OS. @@ -2011,7 +2312,7 @@ def _OSOndiskAPIVersion(os_dir): st = os.stat(api_file) except EnvironmentError, err: return False, ("Required file '%s' not found under path %s: %s" % - (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err))) + (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err))) if not stat.S_ISREG(stat.S_IFMT(st.st_mode)): return False, ("File '%s' in %s is not a regular file" % @@ -2021,7 +2322,7 @@ def _OSOndiskAPIVersion(os_dir): 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_file, utils.ErrnoOrStr(err))) try: api_versions = [int(version.strip()) for version in api_versions] @@ -2038,7 +2339,7 @@ def DiagnoseOS(top_dirs=None): @type top_dirs: list @param top_dirs: the list of directories in which to search (if not given defaults to - L{constants.OS_SEARCH_PATH}) + L{pathutils.OS_SEARCH_PATH}) @rtype: list of L{objects.OS} @return: a list of tuples (name, path, status, diagnose, variants, parameters, api_version) for all (potential) OSes under all @@ -2053,7 +2354,7 @@ def DiagnoseOS(top_dirs=None): """ if top_dirs is None: - top_dirs = constants.OS_SEARCH_PATH + top_dirs = pathutils.OS_SEARCH_PATH result = [] for dir_name in top_dirs: @@ -2095,7 +2396,7 @@ def _TryOSFromDisk(name, base_dir=None): """ if base_dir is None: - os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir) + os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir) else: os_dir = utils.FindFile(name, [base_dir], os.path.isdir) @@ -2134,7 +2435,7 @@ def _TryOSFromDisk(name, base_dir=None): del os_files[filename] continue return False, ("File '%s' under path '%s' is missing (%s)" % - (filename, os_dir, _ErrnoOrStr(err))) + (filename, os_dir, utils.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" % @@ -2149,12 +2450,13 @@ def _TryOSFromDisk(name, base_dir=None): if constants.OS_VARIANTS_FILE in os_files: variants_file = os_files[constants.OS_VARIANTS_FILE] try: - variants = utils.ReadFile(variants_file).splitlines() + variants = \ + utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file)) except EnvironmentError, err: # we accept missing files, but not other errors if err.errno != errno.ENOENT: return False, ("Error while reading the OS variants file at %s: %s" % - (variants_file, _ErrnoOrStr(err))) + (variants_file, utils.ErrnoOrStr(err))) parameters = [] if constants.OS_PARAMETERS_FILE in os_files: @@ -2163,7 +2465,7 @@ def _TryOSFromDisk(name, base_dir=None): parameters = utils.ReadFile(parameters_file).splitlines() except EnvironmentError, err: return False, ("Error while reading the OS parameters file at %s: %s" % - (parameters_file, _ErrnoOrStr(err))) + (parameters_file, utils.ErrnoOrStr(err))) parameters = [v.split(None, 1) for v in parameters] os_obj = objects.OS(name=name, path=os_dir, @@ -2243,6 +2545,11 @@ def OSCoreEnv(os_name, inst_os, os_params, debug=0): for pname, pvalue in os_params.items(): result["OSP_%s" % pname.upper()] = pvalue + # Set a default path otherwise programs called by OS scripts (or + # even hooks called from OS scripts) might break, and we don't want + # to have each script require setting a PATH variable + result["PATH"] = constants.HOOKS_PATH + return result @@ -2277,6 +2584,9 @@ def OSEnvironment(instance, inst_os, debug=0): real_disk = _OpenRealBD(disk) result["DISK_%d_PATH" % idx] = real_disk.dev_path result["DISK_%d_ACCESS" % idx] = disk.mode + result["DISK_%d_UUID" % idx] = disk.uuid + if disk.name: + result["DISK_%d_NAME" % idx] = disk.name if constants.HV_DISK_TYPE in instance.hvparams: result["DISK_%d_FRONTEND_TYPE" % idx] = \ instance.hvparams[constants.HV_DISK_TYPE] @@ -2289,6 +2599,9 @@ def OSEnvironment(instance, inst_os, debug=0): # NICs for idx, nic in enumerate(instance.nics): result["NIC_%d_MAC" % idx] = nic.mac + result["NIC_%d_UUID" % idx] = nic.uuid + if nic.name: + result["NIC_%d_NAME" % idx] = nic.name if nic.ip: result["NIC_%d_IP" % idx] = nic.ip result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE] @@ -2296,6 +2609,9 @@ def OSEnvironment(instance, inst_os, debug=0): result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK] if nic.nicparams[constants.NIC_LINK]: result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK] + if nic.netinfo: + nobj = objects.Network.FromDict(nic.netinfo) + result.update(nobj.HooksDict("NIC_%d_" % idx)) if constants.HV_NIC_TYPE in instance.hvparams: result["NIC_%d_FRONTEND_TYPE" % idx] = \ instance.hvparams[constants.HV_NIC_TYPE] @@ -2308,7 +2624,52 @@ def OSEnvironment(instance, inst_os, debug=0): return result -def BlockdevGrow(disk, amount, dryrun): +def DiagnoseExtStorage(top_dirs=None): + """Compute the validity for all ExtStorage Providers. + + @type top_dirs: list + @param top_dirs: the list of directories in which to + search (if not given defaults to + L{pathutils.ES_SEARCH_PATH}) + @rtype: list of L{objects.ExtStorage} + @return: a list of tuples (name, path, status, diagnose, parameters) + for all (potential) ExtStorage Providers under all + search paths, where: + - name is the (potential) ExtStorage Provider + - path is the full path to the ExtStorage Provider + - status True/False is the validity of the ExtStorage Provider + - diagnose is the error message for an invalid ExtStorage Provider, + otherwise empty + - parameters is a list of (name, help) parameters, if any + + """ + if top_dirs is None: + top_dirs = pathutils.ES_SEARCH_PATH + + result = [] + for dir_name in top_dirs: + if os.path.isdir(dir_name): + try: + f_names = utils.ListVisibleFiles(dir_name) + except EnvironmentError, err: + logging.exception("Can't list the ExtStorage directory %s: %s", + dir_name, err) + break + for name in f_names: + es_path = utils.PathJoin(dir_name, name) + status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name) + if status: + diagnose = "" + parameters = es_inst.supported_parameters + else: + diagnose = es_inst + parameters = [] + result.append((name, es_path, status, diagnose, parameters)) + + return result + + +def BlockdevGrow(disk, amount, dryrun, backingstore): """Grow a stack of block devices. This function is called recursively, with the childrens being the @@ -2321,6 +2682,9 @@ def BlockdevGrow(disk, amount, dryrun): @type dryrun: boolean @param dryrun: whether to execute the operation in simulation mode only, without actually increasing the size + @param backingstore: whether to execute the operation on backing storage + only, or on "logical" storage only; e.g. DRBD is logical storage, + whereas LVM, file, RBD are backing storage @rtype: (status, result) @return: a tuple with the status of the operation (True/False), and the errors message if status is False @@ -2331,7 +2695,7 @@ def BlockdevGrow(disk, amount, dryrun): _Fail("Cannot find block device %s", disk) try: - r_dev.Grow(amount, dryrun) + r_dev.Grow(amount, dryrun, backingstore) except errors.BlockDeviceError, err: _Fail("Failed to grow block device: %s", err, exc=True) @@ -2366,6 +2730,32 @@ def BlockdevSnapshot(disk): disk.unique_id, disk.dev_type) +def BlockdevSetInfo(disk, info): + """Sets 'metadata' information on block devices. + + This function sets 'info' metadata on block devices. Initial + information is set at device creation; this function should be used + for example after renames. + + @type disk: L{objects.Disk} + @param disk: the disk to be grown + @type info: string + @param info: new 'info' metadata + @rtype: (status, result) + @return: a tuple with the status of the operation (True/False), and + the errors message if status is False + + """ + r_dev = _RecursiveFindBD(disk) + if r_dev is None: + _Fail("Cannot find block device %s", disk) + + try: + r_dev.SetInfo(info) + except errors.BlockDeviceError, err: + _Fail("Failed to set information on block device: %s", err, exc=True) + + def FinalizeExport(instance, snap_disks): """Write out the export configuration information. @@ -2379,8 +2769,8 @@ def FinalizeExport(instance, snap_disks): @rtype: None """ - destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new") - finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name) + destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new") + finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name) config = objects.SerializableConfigParser() @@ -2393,8 +2783,13 @@ def FinalizeExport(instance, snap_disks): config.add_section(constants.INISECT_INS) config.set(constants.INISECT_INS, "name", instance.name) + config.set(constants.INISECT_INS, "maxmem", "%d" % + instance.beparams[constants.BE_MAXMEM]) + config.set(constants.INISECT_INS, "minmem", "%d" % + instance.beparams[constants.BE_MINMEM]) + # "memory" is deprecated, but useful for exporting to old ganeti versions config.set(constants.INISECT_INS, "memory", "%d" % - instance.beparams[constants.BE_MEMORY]) + instance.beparams[constants.BE_MAXMEM]) config.set(constants.INISECT_INS, "vcpus", "%d" % instance.beparams[constants.BE_VCPUS]) config.set(constants.INISECT_INS, "disk_template", instance.disk_template) @@ -2407,6 +2802,8 @@ def FinalizeExport(instance, snap_disks): 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_network" % nic_count, + "%s" % nic.network) for param in constants.NICS_PARAMETER_TYPES: config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param), "%s" % nic.nicparams.get(param, None)) @@ -2477,8 +2874,8 @@ def ListExports(): @return: list of the exports """ - if os.path.isdir(constants.EXPORT_DIR): - return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR)) + if os.path.isdir(pathutils.EXPORT_DIR): + return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR)) else: _Fail("No exports directory") @@ -2491,7 +2888,7 @@ def RemoveExport(export): @rtype: None """ - target = utils.PathJoin(constants.EXPORT_DIR, export) + target = utils.PathJoin(pathutils.EXPORT_DIR, export) try: shutil.rmtree(target) @@ -2553,18 +2950,13 @@ def _TransformFileStorageDir(fs_dir): @return: the normalized path if valid, None otherwise """ - if not constants.ENABLE_FILE_STORAGE: + if not (constants.ENABLE_FILE_STORAGE or + constants.ENABLE_SHARED_FILE_STORAGE): _Fail("File storage disabled at configure time") - cfg = _GetConfig() - fs_dir = os.path.normpath(fs_dir) - base_fstore = cfg.GetFileStorageDir() - base_shared = cfg.GetSharedFileStorageDir() - if not (utils.IsBelowDir(base_fstore, fs_dir) or - utils.IsBelowDir(base_shared, fs_dir)): - _Fail("File storage directory '%s' is not under base file" - " storage directory '%s' or shared storage directory '%s'", - fs_dir, base_fstore, base_shared) - return fs_dir + + bdev.CheckFileStoragePath(fs_dir) + + return os.path.normpath(fs_dir) def CreateFileStorageDir(file_storage_dir): @@ -2655,12 +3047,9 @@ def _EnsureJobQueueFile(file_name): @raises RPCFail: if the file is not valid """ - queue_dir = os.path.normpath(constants.QUEUE_DIR) - result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir) - - if not result: + if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name): _Fail("Passed job queue file '%s' does not belong to" - " the queue directory '%s'", file_name, queue_dir) + " the queue directory '%s'", file_name, pathutils.QUEUE_DIR) def JobQueueUpdate(file_name, content): @@ -2677,12 +3066,14 @@ def JobQueueUpdate(file_name, content): @return: the success of the operation """ + file_name = vcluster.LocalizeVirtualPath(file_name) + _EnsureJobQueueFile(file_name) getents = runtime.GetEnts() # Write and replace the file atomically utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid, - gid=getents.masterd_gid) + gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS) def JobQueueRename(old, new): @@ -2698,10 +3089,16 @@ def JobQueueRename(old, new): @return: the success of the operation and payload """ + old = vcluster.LocalizeVirtualPath(old) + new = vcluster.LocalizeVirtualPath(new) + _EnsureJobQueueFile(old) _EnsureJobQueueFile(new) - utils.RenameFile(old, new, mkdir=True) + getents = runtime.GetEnts() + + utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750, + dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid) def BlockdevClose(instance_name, disks): @@ -2831,18 +3228,18 @@ def DemoteFromMC(): if master == myself: _Fail("ssconf status shows I'm the master node, will not demote") - result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD]) + result = utils.RunCmd([pathutils.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) + if os.path.isfile(pathutils.CLUSTER_CONF_FILE): + utils.CreateBackup(pathutils.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) + utils.RemoveFile(pathutils.CLUSTER_CONF_FILE) def _GetX509Filenames(cryptodir, name): @@ -2854,7 +3251,7 @@ def _GetX509Filenames(cryptodir, name): utils.PathJoin(cryptodir, name, _X509_CERT_FILE)) -def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR): +def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR): """Creates a new X509 certificate for SSL/TLS. @type validity: int @@ -2885,7 +3282,7 @@ def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR): raise -def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR): +def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR): """Removes a X509 certificate. @type name: string @@ -2930,9 +3327,9 @@ def _GetImportExportIoCommand(instance, mode, ieio, ieargs): real_filename = os.path.realpath(filename) directory = os.path.dirname(real_filename) - if not utils.IsBelowDir(constants.EXPORT_DIR, real_filename): + if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename): _Fail("File '%s' is not under exports directory '%s': %s", - filename, constants.EXPORT_DIR, real_filename) + filename, pathutils.EXPORT_DIR, real_filename) # Create directory utils.Makedirs(directory, mode=0750) @@ -3019,7 +3416,7 @@ def _CreateImportExportStatusDir(prefix): """Creates status directory for import/export. """ - return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR, + return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR, prefix=("%s-%s-" % (prefix, utils.TimestampForFilename()))) @@ -3067,11 +3464,11 @@ def StartImportExportDaemon(mode, opts, host, port, instance, component, if opts.key_name is None: # Use server.pem - key_path = constants.NODED_CERT_FILE - cert_path = constants.NODED_CERT_FILE + key_path = pathutils.NODED_CERT_FILE + cert_path = pathutils.NODED_CERT_FILE assert opts.ca_pem is None else: - (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR, + (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR, opts.key_name) assert opts.ca_pem is not None @@ -3087,7 +3484,7 @@ def StartImportExportDaemon(mode, opts, host, port, instance, component, if opts.ca_pem is None: # Use server.pem - ca = utils.ReadFile(constants.NODED_CERT_FILE) + ca = utils.ReadFile(pathutils.NODED_CERT_FILE) else: ca = opts.ca_pem @@ -3095,7 +3492,7 @@ def StartImportExportDaemon(mode, opts, host, port, instance, component, utils.WriteFile(ca_file, data=ca, mode=0400) cmd = [ - constants.IMPORT_EXPORT_DAEMON, + pathutils.IMPORT_EXPORT_DAEMON, status_file, mode, "--key=%s" % key_path, "--cert=%s" % cert_path, @@ -3165,7 +3562,7 @@ def GetImportExportStatus(names): result = [] for name in names: - status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name, + status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name, _IES_STATUS_FILE) try: @@ -3190,7 +3587,7 @@ def AbortImportExport(name): """ logging.info("Abort import/export %s", name) - status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name) + status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name) pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE)) if pid: @@ -3208,7 +3605,7 @@ def CleanupImportExport(name): """ logging.info("Finalizing import/export %s", name) - status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name) + status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name) pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE)) @@ -3286,8 +3683,20 @@ def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster): for rd in bdevs: stats = rd.GetProcStatus() - all_connected = (all_connected and - (stats.is_connected or stats.is_in_resync)) + if multimaster: + # In the multimaster case we have to wait explicitly until + # the resource is Connected and UpToDate/UpToDate, because + # we promote *both nodes* to primary directly afterwards. + # Being in resync is not enough, since there is a race during which we + # may promote a node with an Outdated disk to primary, effectively + # tearing down the connection. + all_connected = (all_connected and + stats.is_connected and + stats.is_disk_uptodate and + stats.peer_disk_uptodate) + else: + 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 @@ -3380,6 +3789,209 @@ def PowercycleNode(hypervisor_type): hyper.PowercycleNode() +def _VerifyRestrictedCmdName(cmd): + """Verifies a restricted command name. + + @type cmd: string + @param cmd: Command name + @rtype: tuple; (boolean, string or None) + @return: The tuple's first element is the status; if C{False}, the second + element is an error message string, otherwise it's C{None} + + """ + if not cmd.strip(): + return (False, "Missing command name") + + if os.path.basename(cmd) != cmd: + return (False, "Invalid command name") + + if not constants.EXT_PLUGIN_MASK.match(cmd): + return (False, "Command name contains forbidden characters") + + return (True, None) + + +def _CommonRestrictedCmdCheck(path, owner): + """Common checks for restricted command file system directories and files. + + @type path: string + @param path: Path to check + @param owner: C{None} or tuple containing UID and GID + @rtype: tuple; (boolean, string or C{os.stat} result) + @return: The tuple's first element is the status; if C{False}, the second + element is an error message string, otherwise it's the result of C{os.stat} + + """ + if owner is None: + # Default to root as owner + owner = (0, 0) + + try: + st = os.stat(path) + except EnvironmentError, err: + return (False, "Can't stat(2) '%s': %s" % (path, err)) + + if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE): + return (False, "Permissions on '%s' are too permissive" % path) + + if (st.st_uid, st.st_gid) != owner: + (owner_uid, owner_gid) = owner + return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid)) + + return (True, st) + + +def _VerifyRestrictedCmdDirectory(path, _owner=None): + """Verifies restricted command directory. + + @type path: string + @param path: Path to check + @rtype: tuple; (boolean, string or None) + @return: The tuple's first element is the status; if C{False}, the second + element is an error message string, otherwise it's C{None} + + """ + (status, value) = _CommonRestrictedCmdCheck(path, _owner) + + if not status: + return (False, value) + + if not stat.S_ISDIR(value.st_mode): + return (False, "Path '%s' is not a directory" % path) + + return (True, None) + + +def _VerifyRestrictedCmd(path, cmd, _owner=None): + """Verifies a whole restricted command and returns its executable filename. + + @type path: string + @param path: Directory containing restricted commands + @type cmd: string + @param cmd: Command name + @rtype: tuple; (boolean, string) + @return: The tuple's first element is the status; if C{False}, the second + element is an error message string, otherwise the second element is the + absolute path to the executable + + """ + executable = utils.PathJoin(path, cmd) + + (status, msg) = _CommonRestrictedCmdCheck(executable, _owner) + + if not status: + return (False, msg) + + if not utils.IsExecutable(executable): + return (False, "access(2) thinks '%s' can't be executed" % executable) + + return (True, executable) + + +def _PrepareRestrictedCmd(path, cmd, + _verify_dir=_VerifyRestrictedCmdDirectory, + _verify_name=_VerifyRestrictedCmdName, + _verify_cmd=_VerifyRestrictedCmd): + """Performs a number of tests on a restricted command. + + @type path: string + @param path: Directory containing restricted commands + @type cmd: string + @param cmd: Command name + @return: Same as L{_VerifyRestrictedCmd} + + """ + # Verify the directory first + (status, msg) = _verify_dir(path) + if status: + # Check command if everything was alright + (status, msg) = _verify_name(cmd) + + if not status: + return (False, msg) + + # Check actual executable + return _verify_cmd(path, cmd) + + +def RunRestrictedCmd(cmd, + _lock_timeout=_RCMD_LOCK_TIMEOUT, + _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE, + _path=pathutils.RESTRICTED_COMMANDS_DIR, + _sleep_fn=time.sleep, + _prepare_fn=_PrepareRestrictedCmd, + _runcmd_fn=utils.RunCmd, + _enabled=constants.ENABLE_RESTRICTED_COMMANDS): + """Executes a restricted command after performing strict tests. + + @type cmd: string + @param cmd: Command name + @rtype: string + @return: Command output + @raise RPCFail: In case of an error + + """ + logging.info("Preparing to run restricted command '%s'", cmd) + + if not _enabled: + _Fail("Restricted commands disabled at configure time") + + lock = None + try: + cmdresult = None + try: + lock = utils.FileLock.Open(_lock_file) + lock.Exclusive(blocking=True, timeout=_lock_timeout) + + (status, value) = _prepare_fn(_path, cmd) + + if status: + cmdresult = _runcmd_fn([value], env={}, reset_env=True, + postfork_fn=lambda _: lock.Unlock()) + else: + logging.error(value) + except Exception: # pylint: disable=W0703 + # Keep original error in log + logging.exception("Caught exception") + + if cmdresult is None: + logging.info("Sleeping for %0.1f seconds before returning", + _RCMD_INVALID_DELAY) + _sleep_fn(_RCMD_INVALID_DELAY) + + # Do not include original error message in returned error + _Fail("Executing command '%s' failed" % cmd) + elif cmdresult.failed or cmdresult.fail_reason: + _Fail("Restricted command '%s' failed: %s; output: %s", + cmd, cmdresult.fail_reason, cmdresult.output) + else: + return cmdresult.output + finally: + if lock is not None: + # Release lock at last + lock.Close() + lock = None + + +def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE): + """Creates or removes the watcher pause file. + + @type until: None or number + @param until: Unix timestamp saying until when the watcher shouldn't run + + """ + if until is None: + logging.info("Received request to no longer pause watcher") + utils.RemoveFile(_filename) + else: + logging.info("Received request to pause watcher until %s", until) + + if not ht.TNumber(until): + _Fail("Duration must be numeric") + + utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644) + + class HooksRunner(object): """Hook runner. @@ -3392,15 +4004,29 @@ class HooksRunner(object): @type hooks_base_dir: str or None @param hooks_base_dir: if not None, this overrides the - L{constants.HOOKS_BASE_DIR} (useful for unittests) + L{pathutils.HOOKS_BASE_DIR} (useful for unittests) """ if hooks_base_dir is None: - hooks_base_dir = constants.HOOKS_BASE_DIR + hooks_base_dir = pathutils.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=C0103 + def RunLocalHooks(self, node_list, hpath, phase, env): + """Check that the hooks will be run only locally and then run them. + + """ + assert len(node_list) == 1 + node = node_list[0] + _, myself = ssconf.GetMasterAndMyself() + assert node == myself + + results = self.RunHooks(hpath, phase, env) + + # Return values in the form expected by HooksMaster + return {node: (None, False, results)} + def RunHooks(self, hpath, phase, env): """Run the scripts in the hooks directory. @@ -3442,7 +4068,7 @@ class HooksRunner(object): runparts_results = utils.RunParts(dir_name, env=env, reset_env=True) - for (relname, relstatus, runresult) in runparts_results: + for (relname, relstatus, runresult) in runparts_results: if relstatus == constants.RUNPARTS_SKIP: rrval = constants.HKR_SKIP output = "" @@ -3506,7 +4132,7 @@ class DevCacheManager(object): """ _DEV_PREFIX = "/dev/" - _ROOT_DIR = constants.BDEV_CACHE_DIR + _ROOT_DIR = pathutils.BDEV_CACHE_DIR @classmethod def _ConvertPath(cls, dev_path):