def config_sync(env, skip_offline_nodes=False):
"""
Send specified local booth configuration to all nodes in cluster.
env -- LibraryEnvironment
skip_offline_nodes -- if True offline nodes will be skipped
"""
config = env.booth.get_config_content()
authfile_path = config_structure.get_authfile(parse(config))
authfile_content = config_files.read_authfile(
env.report_processor, authfile_path
)
cluster_nodes_names, report_list = get_existing_nodes_names(
env.get_corosync_conf()
)
if not cluster_nodes_names:
report_list.append(reports.corosync_config_no_nodes_defined())
env.report_processor.process_list(report_list)
com_cmd = BoothSendConfig(
env.report_processor,
env.booth.name,
config,
authfile=authfile_path,
authfile_data=authfile_content,
skip_offline_targets=skip_offline_nodes
)
com_cmd.set_targets(
env.get_node_target_factory().get_target_list(
cluster_nodes_names,
skip_non_existing=skip_offline_nodes,
)
)
run_and_raise(env.get_node_communicator(), com_cmd)
def get_cluster_sbd_config(lib_env):
"""
Returns list of SBD config from all cluster nodes in cluster. Structure
of data:
[
{
"node": <NodeAddress>
"config": <sbd_config_dict> or None if there was failure,
},
...
]
If error occurs while obtaining config from some node, it's config will be
None. If obtaining config fail on all node returns empty dictionary.
lib_env -- LibraryEnvironment
"""
node_list, get_nodes_report_list = get_existing_nodes_names(
lib_env.get_corosync_conf()
)
if not node_list:
get_nodes_report_list.append(reports.corosync_config_no_nodes_defined())
lib_env.report_processor.process_list(get_nodes_report_list)
com_cmd = GetSbdConfig(lib_env.report_processor)
com_cmd.set_targets(
lib_env.get_node_target_factory().get_target_list(
node_list,
skip_non_existing=True
)
)
return run_com(lib_env.get_node_communicator(), com_cmd)
def push_corosync_conf(
self, corosync_conf_facade, skip_offline_nodes=False
):
corosync_conf_data = corosync_conf_facade.config.export()
if self.is_corosync_conf_live:
node_name_list, report_list = get_existing_nodes_names(
corosync_conf_facade,
# Pcs is unable to communicate with nodes missing names. It
# cannot send new corosync.conf to them. That might break the
# cluster. Hence we error out.
error_on_missing_name=True
)
self.report_processor.process_list(report_list)
self._push_corosync_conf_live(
self.get_node_target_factory().get_target_list(
node_name_list,
skip_non_existing=skip_offline_nodes,
),
corosync_conf_data,
corosync_conf_facade.need_stopped_cluster,
corosync_conf_facade.need_qdevice_reload,
skip_offline_nodes,
)
else:
self._corosync_conf_data = corosync_conf_data
def get_cluster_sbd_status(lib_env):
"""
Returns status of SBD service in cluster in dictionary with format:
{
<NodeAddress>: {
"installed": <boolean>,
"enabled": <boolean>,
"running": <boolean>
},
...
}
lib_env -- LibraryEnvironment
"""
node_list, get_nodes_report_list = get_existing_nodes_names(
lib_env.get_corosync_conf()
)
if not node_list:
get_nodes_report_list.append(reports.corosync_config_no_nodes_defined())
lib_env.report_processor.process_list(get_nodes_report_list)
com_cmd = GetSbdStatus(lib_env.report_processor)
com_cmd.set_targets(
lib_env.get_node_target_factory().get_target_list(
node_list,
skip_non_existing=True
)
)
return run_com(lib_env.get_node_communicator(), com_cmd)
def quorum_unblock_cmd(lib, argv, modifiers):
"""
Options:
* --force - no error when removing non existing property and no warning
about this action
"""
modifiers.ensure_only_supported("--force")
if argv:
raise CmdLineInputError()
output, retval = utils.run(
["corosync-cmapctl", "-g", "runtime.votequorum.wait_for_all_status"]
)
if retval != 0:
utils.err("unable to check quorum status")
if output.split("=")[-1].strip() != "1":
utils.err("cluster is not waiting for nodes to establish quorum")
all_nodes, report_list = get_existing_nodes_names(
utils.get_corosync_conf_facade()
)
if report_list:
utils.process_library_reports(report_list)
unjoined_nodes = set(all_nodes) - set(utils.getCorosyncActiveNodes())
if not unjoined_nodes:
utils.err("no unjoined nodes found")
if not modifiers.get("--force"):
answer = utils.get_terminal_input(
(
"WARNING: If node(s) {nodes} are not powered off or they do"
+ " have access to shared resources, data corruption and/or"
+ " cluster failure may occur. Are you sure you want to"
+ " continue? [y/N] "
).format(nodes=", ".join(unjoined_nodes))
)
if answer.lower() not in ["y", "yes"]:
print("Canceled")
return
for node in unjoined_nodes:
# pass --force so no warning will be displayed
stonith.stonith_confirm(
lib, [node], parse_args.InputModifiers({"--force": ""})
)
output, retval = utils.run(
["corosync-cmapctl", "-s", "quorum.cancel_wait_for_all", "u8", "1"]
)
if retval != 0:
utils.err("unable to cancel waiting for nodes")
print("Quorum unblocked")
startup_fencing = utils.get_set_properties().get("startup-fencing", "")
utils.set_cib_property(
"startup-fencing",
"false" if startup_fencing.lower() != "false" else "true"
)
utils.set_cib_property("startup-fencing", startup_fencing)
print("Waiting for nodes canceled")
def disable_sbd(lib_env, ignore_offline_nodes=False):
"""
Disable SBD on all nodes in cluster.
lib_env -- LibraryEnvironment
ignore_offline_nodes -- if True, omit offline nodes
"""
node_list, get_nodes_report_list = get_existing_nodes_names(
lib_env.get_corosync_conf()
)
if not node_list:
get_nodes_report_list.append(reports.corosync_config_no_nodes_defined())
lib_env.report_processor.process_list(get_nodes_report_list)
com_cmd = GetOnlineTargets(
lib_env.report_processor, ignore_offline_targets=ignore_offline_nodes,
)
com_cmd.set_targets(
lib_env.get_node_target_factory().get_target_list(
node_list,
skip_non_existing=ignore_offline_nodes,
)
)
online_nodes = run_and_raise(lib_env.get_node_communicator(), com_cmd)
com_cmd = SetStonithWatchdogTimeoutToZero(lib_env.report_processor)
com_cmd.set_targets(online_nodes)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
com_cmd = DisableSbdService(lib_env.report_processor)
com_cmd.set_targets(online_nodes)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
lib_env.report_processor.process(
reports.cluster_restart_required_to_apply_changes()
)
def nodes_status(lib, argv, modifiers):
"""
Options:
* -f - CIB file - for config subcommand and not for both or corosync
* --corosync_conf - only for config subcommand
NOTE: modifiers check is in subcommand
"""
del lib
if len(argv) == 1 and (argv[0] == "config"):
modifiers.ensure_only_supported("-f", "--corosync_conf")
if utils.hasCorosyncConf():
corosync_nodes, report_list = get_existing_nodes_names(
utils.get_corosync_conf_facade())
if report_list:
process_library_reports(report_list)
else:
corosync_nodes = []
try:
pacemaker_nodes = sorted([
node.attrs.name for node in ClusterState(
get_cluster_status_dom(
utils.cmd_runner())).node_section.nodes
if node.attrs.type != "remote"
])
except LibraryError as e:
process_library_reports(e.args)
print("Corosync Nodes:")
if corosync_nodes:
print(" " + " ".join(corosync_nodes))
print("Pacemaker Nodes:")
if pacemaker_nodes:
print(" " + " ".join(pacemaker_nodes))
return
if len(argv) == 1 and (argv[0] == "corosync" or argv[0] == "both"):
modifiers.ensure_only_supported()
all_nodes, report_list = get_existing_nodes_names(
utils.get_corosync_conf_facade())
if report_list:
process_library_reports(report_list)
online_nodes = utils.getCorosyncActiveNodes()
offline_nodes = []
for node in all_nodes:
if node not in online_nodes:
offline_nodes.append(node)
online_nodes.sort()
offline_nodes.sort()
print("Corosync Nodes:")
print(" ".join([" Online:"] + online_nodes))
print(" ".join([" Offline:"] + offline_nodes))
if argv[0] != "both":
sys.exit(0)
modifiers.ensure_only_supported("-f")
info_dom = utils.getClusterState()
nodes = info_dom.getElementsByTagName("nodes")
if nodes.length == 0:
utils.err("No nodes section found")
onlinenodes = []
offlinenodes = []
standbynodes = []
standbynodes_with_resources = []
maintenancenodes = []
remote_onlinenodes = []
remote_offlinenodes = []
remote_standbynodes = []
remote_standbynodes_with_resources = []
remote_maintenancenodes = []
for node in nodes[0].getElementsByTagName("node"):
node_name = node.getAttribute("name")
node_remote = node.getAttribute("type") == "remote"
if node.getAttribute("online") == "true":
if node.getAttribute("standby") == "true":
is_running_resources = (node.getAttribute("resources_running")
!= "0")
if node_remote:
if is_running_resources:
remote_standbynodes_with_resources.append(node_name)
else:
remote_standbynodes.append(node_name)
else:
if is_running_resources:
standbynodes_with_resources.append(node_name)
else:
standbynodes.append(node_name)
if node.getAttribute("maintenance") == "true":
if node_remote:
remote_maintenancenodes.append(node_name)
else:
maintenancenodes.append(node_name)
if (node.getAttribute("standby") == "false"
and node.getAttribute("maintenance") == "false"):
if node_remote:
remote_onlinenodes.append(node_name)
else:
onlinenodes.append(node_name)
else:
if node_remote:
remote_offlinenodes.append(node_name)
else:
offlinenodes.append(node_name)
print("Pacemaker Nodes:")
print(" ".join([" Online:"] + onlinenodes))
print(" ".join([" Standby:"] + standbynodes))
print(" ".join([" Standby with resource(s) running:"] +
standbynodes_with_resources))
print(" ".join([" Maintenance:"] + maintenancenodes))
print(" ".join([" Offline:"] + offlinenodes))
print("Pacemaker Remote Nodes:")
print(" ".join([" Online:"] + remote_onlinenodes))
print(" ".join([" Standby:"] + remote_standbynodes))
print(" ".join([" Standby with resource(s) running:"] +
remote_standbynodes_with_resources))
print(" ".join([" Maintenance:"] + remote_maintenancenodes))
print(" ".join([" Offline:"] + remote_offlinenodes))
def status_all_sites_plaintext(
env: LibraryEnvironment,
hide_inactive_resources: bool = False,
verbose: bool = False,
) -> List[Mapping[str, Any]]:
"""
Return local site's and all remote sites' status as plaintext
env -- LibraryEnvironment
hide_inactive_resources -- if True, do not display non-running resources
verbose -- if True, display more info
"""
# The command does not provide an option to skip offline / unreacheable /
# misbehaving nodes.
# The point of such skipping is to stop a command if it is unable to make
# changes on all nodes. The user can then decide to proceed anyway and
# make changes on the skipped nodes later manually.
# This command only reads from nodes so it automatically asks other nodes
# if one is offline / misbehaving.
class SiteData():
def __init__(
self,
local: bool,
role: DrRole,
target_list: Iterable[RequestTarget],
) -> None:
self.local = local
self.role = role
self.target_list = target_list
self.status_loaded = False
self.status_plaintext = ""
if env.ghost_file_codes:
raise LibraryError(
reports.live_environment_required(env.ghost_file_codes))
report_processor = env.report_processor
report_list, dr_config = _load_dr_config(env.get_dr_env().config)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
site_data_list = []
target_factory = env.get_node_target_factory()
# get local nodes
local_nodes, report_list = get_existing_nodes_names(
env.get_corosync_conf())
report_processor.report_list(report_list)
report_list, local_targets = target_factory.get_target_list_with_reports(
local_nodes,
skip_non_existing=True,
)
report_processor.report_list(report_list)
site_data_list.append(SiteData(True, dr_config.local_role, local_targets))
# get remote sites' nodes
for conf_remote_site in dr_config.get_remote_site_list():
report_list, remote_targets = (
target_factory.get_target_list_with_reports(
conf_remote_site.node_name_list,
skip_non_existing=True,
))
report_processor.report_list(report_list)
site_data_list.append(
SiteData(False, conf_remote_site.role, remote_targets))
if report_processor.has_errors:
raise LibraryError()
# get all statuses
for site_data in site_data_list:
com_cmd = GetFullClusterStatusPlaintext(
report_processor,
hide_inactive_resources=hide_inactive_resources,
verbose=verbose,
)
com_cmd.set_targets(site_data.target_list)
site_data.status_loaded, site_data.status_plaintext = run_com_cmd(
env.get_node_communicator(), com_cmd)
return [
dto.to_dict(
DrSiteStatusDto(
local_site=site_data.local,
site_role=site_data.role,
status_plaintext=site_data.status_plaintext,
status_successfully_obtained=site_data.status_loaded,
)) for site_data in site_data_list
]
def add_device(
lib_env, model, model_options, generic_options, heuristics_options,
force_model=False, force_options=False, skip_offline_nodes=False
):
"""
Add a quorum device to a cluster, distribute and reload configs if live
string model -- quorum device model
dict model_options -- model specific options
dict generic_options -- generic quorum device options
dict heuristics_options -- heuristics options
bool force_model -- continue even if the model is not valid
bool force_options -- continue even if options are not valid
bool skip_offline_nodes -- continue even if not all nodes are accessible
"""
cfg = lib_env.get_corosync_conf()
if cfg.has_quorum_device():
raise LibraryError(reports.qdevice_already_defined())
report_processor = SimpleReportProcessor(lib_env.report_processor)
report_processor.report_list(
corosync_conf_validators.add_quorum_device(
model,
model_options,
generic_options,
heuristics_options,
[node.nodeid for node in cfg.get_nodes()],
force_model=force_model,
force_options=force_options
)
)
if lib_env.is_corosync_conf_live:
cluster_nodes_names, report_list = get_existing_nodes_names(
cfg,
# Pcs is unable to communicate with nodes missing names. It cannot
# send new corosync.conf to them. That might break the cluster.
# Hence we error out.
error_on_missing_name=True
)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
cfg.add_quorum_device(
model,
model_options,
generic_options,
heuristics_options,
)
if cfg.is_quorum_device_heuristics_enabled_with_no_exec():
lib_env.report_processor.process(
reports.corosync_quorum_heuristics_enabled_with_no_exec()
)
# First setup certificates for qdevice, then send corosync.conf to nodes.
# If anything fails, nodes will not have corosync.conf with qdevice in it,
# so there is no effect on the cluster.
if lib_env.is_corosync_conf_live:
target_factory = lib_env.get_node_target_factory()
target_list = target_factory.get_target_list(
cluster_nodes_names, skip_non_existing=skip_offline_nodes,
)
# Do model specific configuration.
# If the model is not known to pcs and was forced, do not configure
# anything else than corosync.conf, as we do not know what to do
# anyway.
if model == "net":
qdevice_net.set_up_client_certificates(
lib_env.cmd_runner(),
lib_env.report_processor,
lib_env.communicator_factory,
# We are sure the "host" key is there, it has been validated
# above.
target_factory.get_target_from_hostname(model_options["host"]),
cfg.get_cluster_name(),
target_list,
skip_offline_nodes
)
lib_env.report_processor.process(
reports.service_enable_started("corosync-qdevice")
)
com_cmd = qdevice_com.Enable(
lib_env.report_processor, skip_offline_nodes
)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# everything set up, it's safe to tell the nodes to use qdevice
lib_env.push_corosync_conf(cfg, skip_offline_nodes)
# Now, when corosync.conf has been reloaded, we can start qdevice service.
if lib_env.is_corosync_conf_live:
lib_env.report_processor.process(
reports.service_start_started("corosync-qdevice")
)
com_cmd = qdevice_com.Start(
lib_env.report_processor, skip_offline_nodes
)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
def remove_device(lib_env, skip_offline_nodes=False):
"""
Stop using quorum device, distribute and reload configs if live
skip_offline_nodes continue even if not all nodes are accessible
"""
cfg = lib_env.get_corosync_conf()
if not cfg.has_quorum_device():
raise LibraryError(reports.qdevice_not_defined())
model = cfg.get_quorum_device_model()
cfg.remove_quorum_device()
if lib_env.is_corosync_conf_live:
report_processor = SimpleReportProcessor(lib_env.report_processor)
# get nodes for communication
cluster_nodes_names, report_list = get_existing_nodes_names(
cfg,
# Pcs is unable to communicate with nodes missing names. It cannot
# send new corosync.conf to them. That might break the cluster.
# Hence we error out.
error_on_missing_name=True
)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
target_list = lib_env.get_node_target_factory().get_target_list(
cluster_nodes_names, skip_non_existing=skip_offline_nodes,
)
# fix quorum options for SBD to work properly
if sbd.atb_has_to_be_enabled(lib_env.cmd_runner(), cfg):
lib_env.report_processor.process(
reports.corosync_quorum_atb_will_be_enabled_due_to_sbd()
)
cfg.set_quorum_options({"auto_tie_breaker": "1"})
# disable qdevice
lib_env.report_processor.process(
reports.service_disable_started("corosync-qdevice")
)
com_cmd = qdevice_com.Disable(
lib_env.report_processor, skip_offline_nodes
)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# stop qdevice
lib_env.report_processor.process(
reports.service_stop_started("corosync-qdevice")
)
com_cmd = qdevice_com.Stop(
lib_env.report_processor, skip_offline_nodes
)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# handle model specific configuration
if model == "net":
lib_env.report_processor.process(
reports.qdevice_certificate_removal_started()
)
com_cmd = qdevice_net_com.ClientDestroy(
lib_env.report_processor, skip_offline_nodes
)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
lib_env.push_corosync_conf(cfg, skip_offline_nodes)
def remove_device(lib_env: LibraryEnvironment, skip_offline_nodes=False):
"""
Stop using quorum device, distribute and reload configs if live
skip_offline_nodes continue even if not all nodes are accessible
"""
cfg = lib_env.get_corosync_conf()
if not cfg.has_quorum_device():
raise LibraryError(
ReportItem.error(reports.messages.QdeviceNotDefined()))
model = cfg.get_quorum_device_model()
cfg.remove_quorum_device()
if lib_env.is_corosync_conf_live:
report_processor = lib_env.report_processor
# get nodes for communication
cluster_nodes_names, report_list = get_existing_nodes_names(
cfg,
# Pcs is unable to communicate with nodes missing names. It cannot
# send new corosync.conf to them. That might break the cluster.
# Hence we error out.
error_on_missing_name=True,
)
if report_processor.report_list(report_list).has_errors:
raise LibraryError()
target_list = lib_env.get_node_target_factory().get_target_list(
cluster_nodes_names,
skip_non_existing=skip_offline_nodes,
)
# fix quorum options for SBD to work properly
if sbd.atb_has_to_be_enabled(lib_env.service_manager, cfg):
lib_env.report_processor.report(
ReportItem.warning(
reports.messages.CorosyncQuorumAtbWillBeEnabledDueToSbd()))
cfg.set_quorum_options({"auto_tie_breaker": "1"})
# disable qdevice
lib_env.report_processor.report(
ReportItem.info(
reports.messages.ServiceActionStarted(
reports.const.SERVICE_ACTION_DISABLE, "corosync-qdevice")))
com_cmd_disable = qdevice_com.Disable(lib_env.report_processor,
skip_offline_nodes)
com_cmd_disable.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd_disable)
# stop qdevice
lib_env.report_processor.report(
ReportItem.info(
reports.messages.ServiceActionStarted(
reports.const.SERVICE_ACTION_STOP, "corosync-qdevice")))
com_cmd_stop = qdevice_com.Stop(lib_env.report_processor,
skip_offline_nodes)
com_cmd_stop.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd_stop)
# handle model specific configuration
if model == "net":
lib_env.report_processor.report(
ReportItem.info(
reports.messages.QdeviceCertificateRemovalStarted()))
com_cmd_client_destroy = qdevice_net_com.ClientDestroy(
lib_env.report_processor, skip_offline_nodes)
com_cmd_client_destroy.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(),
com_cmd_client_destroy)
lib_env.push_corosync_conf(cfg, skip_offline_nodes)
def full_cluster_status_plaintext(
env: LibraryEnvironment,
hide_inactive_resources: bool = False,
verbose: bool = False,
) -> str:
"""
Return full cluster status as plaintext
env -- LibraryEnvironment
hide_inactive_resources -- if True, do not display non-running resources
verbose -- if True, display more info
"""
# pylint: disable=too-many-branches
# pylint: disable=too-many-locals
# validation
if not env.is_cib_live and env.is_corosync_conf_live:
raise LibraryError(
reports.live_environment_not_consistent(
[file_type_codes.CIB],
[file_type_codes.COROSYNC_CONF],
))
if env.is_cib_live and not env.is_corosync_conf_live:
raise LibraryError(
reports.live_environment_not_consistent(
[file_type_codes.COROSYNC_CONF],
[file_type_codes.CIB],
))
# initialization
runner = env.cmd_runner()
report_processor = SimpleReportProcessor(env.report_processor)
live = env.is_cib_live and env.is_corosync_conf_live
is_sbd_running = False
# load status, cib, corosync.conf
status_text, warning_list = get_cluster_status_text(
runner, hide_inactive_resources, verbose)
corosync_conf = env.get_corosync_conf()
cib = env.get_cib()
if verbose:
ticket_status_text, ticket_status_stderr, ticket_status_retval = (
get_ticket_status_text(runner))
# get extra info if live
if live:
try:
is_sbd_running = is_service_running(runner, get_sbd_service_name())
except LibraryError:
pass
local_services_status = _get_local_services_status(runner)
if verbose:
node_name_list, node_names_report_list = get_existing_nodes_names(
corosync_conf)
report_processor.report_list(node_names_report_list)
node_reachability = _get_node_reachability(
env.get_node_target_factory(),
env.get_node_communicator(),
report_processor,
node_name_list,
)
# check stonith configuration
warning_list = list(warning_list)
warning_list.extend(_stonith_warnings(cib, is_sbd_running))
# put it all together
if report_processor.has_errors:
raise LibraryError()
parts = []
parts.append(f"Cluster name: {corosync_conf.get_cluster_name()}")
if warning_list:
parts.extend(["", "WARNINGS:"] + warning_list + [""])
parts.append(status_text)
if verbose:
parts.extend(["", "Tickets:"])
if ticket_status_retval != 0:
ticket_warning_parts = [
"WARNING: Unable to get information about tickets"
]
if ticket_status_stderr:
ticket_warning_parts.extend(
indent(ticket_status_stderr.splitlines()))
parts.extend(indent(ticket_warning_parts))
else:
parts.extend(indent(ticket_status_text.splitlines()))
if live:
if verbose:
parts.extend(["", "PCSD Status:"])
parts.extend(
indent(
_format_node_reachability(node_name_list,
node_reachability)))
parts.extend(["", "Daemon Status:"])
parts.extend(
indent(_format_local_services_status(local_services_status)))
return "\n".join(parts)
def set_recovery_site(env: LibraryEnvironment, node_name: str) -> None:
"""
Set up disaster recovery with the local cluster being the primary site
env
node_name -- a known host from the recovery site
"""
# pylint: disable=too-many-locals
if env.ghost_file_codes:
raise LibraryError(
ReportItem.error(
reports.messages.LiveEnvironmentRequired(
env.ghost_file_codes)))
report_processor = env.report_processor
dr_env = env.get_dr_env()
if dr_env.config.raw_file.exists():
report_processor.report(
ReportItem.error(reports.messages.DrConfigAlreadyExist()))
target_factory = env.get_node_target_factory()
local_nodes, report_list = get_existing_nodes_names(
env.get_corosync_conf(), error_on_missing_name=True)
report_processor.report_list(report_list)
if node_name in local_nodes:
report_processor.report(
ReportItem.error(reports.messages.NodeInLocalCluster(node_name)))
report_list, local_targets = target_factory.get_target_list_with_reports(
local_nodes, allow_skip=False, report_none_host_found=False)
report_processor.report_list(report_list)
report_list, remote_targets = target_factory.get_target_list_with_reports(
[node_name], allow_skip=False, report_none_host_found=False)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
# TODO The new file framework doesn't support network communication yet.
com_cmd = GetCorosyncConf(env.report_processor)
com_cmd.set_targets(remote_targets)
corosync_conf_instance = FileInstance.for_corosync_conf()
try:
remote_cluster_nodes, report_list = get_existing_nodes_names(
cast(
CorosyncConfigFacade,
corosync_conf_instance.raw_to_facade(
run_and_raise(env.get_node_communicator(),
com_cmd).encode("utf-8")),
),
error_on_missing_name=True,
)
except ParserErrorException as e:
report_processor.report_list(
corosync_conf_instance.toolbox.parser.exception_to_report_list(
e,
file_type_codes.COROSYNC_CONF,
None,
force_code=None,
is_forced_or_warning=False,
))
if report_processor.report_list(report_list).has_errors:
raise LibraryError()
# ensure we have tokens for all nodes of remote cluster
report_list, remote_targets = target_factory.get_target_list_with_reports(
remote_cluster_nodes, allow_skip=False, report_none_host_found=False)
if report_processor.report_list(report_list).has_errors:
raise LibraryError()
dr_config_exporter = get_file_toolbox(
file_type_codes.PCS_DR_CONFIG).exporter
# create dr config for remote cluster
remote_dr_cfg = dr_env.create_facade(DrRole.RECOVERY)
remote_dr_cfg.add_site(DrRole.PRIMARY, local_nodes)
# send config to all node of remote cluster
distribute_file_cmd = DistributeFilesWithoutForces(
env.report_processor,
node_communication_format.pcs_dr_config_file(
dr_config_exporter.export(remote_dr_cfg.config)),
)
distribute_file_cmd.set_targets(remote_targets)
run_and_raise(env.get_node_communicator(), distribute_file_cmd)
# create new dr config, with local cluster as primary site
local_dr_cfg = dr_env.create_facade(DrRole.PRIMARY)
local_dr_cfg.add_site(DrRole.RECOVERY, remote_cluster_nodes)
distribute_file_cmd = DistributeFilesWithoutForces(
env.report_processor,
node_communication_format.pcs_dr_config_file(
dr_config_exporter.export(local_dr_cfg.config)),
)
distribute_file_cmd.set_targets(local_targets)
run_and_raise(env.get_node_communicator(), distribute_file_cmd)
def location_prefer(lib, argv, modifiers):
"""
Options:
* --force - allow unknown options, allow constraint for any resource type
* -f - CIB file
"""
modifiers.ensure_only_supported("--force", "-f")
rsc = argv.pop(0)
prefer_option = argv.pop(0)
dummy_rsc_type, rsc_value = parse_args.parse_typed_arg(
rsc, [RESOURCE_TYPE_RESOURCE, RESOURCE_TYPE_REGEXP],
RESOURCE_TYPE_RESOURCE)
if prefer_option == "prefers":
prefer = True
elif prefer_option == "avoids":
prefer = False
else:
raise CmdLineInputError()
skip_node_check = False
if modifiers.is_specified("-f") or modifiers.get("--force"):
skip_node_check = True
warn(LOCATION_NODE_VALIDATION_SKIP_MSG)
else:
lib_env = utils.get_lib_env()
existing_nodes, report_list = get_existing_nodes_names(
corosync_conf=lib_env.get_corosync_conf(),
cib=lib_env.get_cib(),
)
if report_list:
process_library_reports(report_list)
report_list = []
parameters_list = []
for nodeconf in argv:
nodeconf_a = nodeconf.split("=", 1)
node = nodeconf_a[0]
if not skip_node_check:
report_list += _verify_node_name(node, existing_nodes)
if len(nodeconf_a) == 1:
if prefer:
score = "INFINITY"
else:
score = "-INFINITY"
else:
score = nodeconf_a[1]
_verify_score(score)
if not prefer:
if score[0] == "-":
score = score[1:]
else:
score = "-" + score
parameters_list.append([
sanitize_id(f"location-{rsc_value}-{node}-{score}"), rsc, node,
score
])
if report_list:
process_library_reports(report_list)
modifiers = modifiers.get_subset("--force", "-f")
for parameters in parameters_list:
location_add(lib,
parameters,
modifiers,
skip_score_and_node_check=True)
def full_cluster_status_plaintext(
env: LibraryEnvironment,
hide_inactive_resources: bool = False,
verbose: bool = False,
) -> str:
"""
Return full cluster status as plaintext
env -- LibraryEnvironment
hide_inactive_resources -- if True, do not display non-running resources
verbose -- if True, display more info
"""
# pylint: disable=too-many-branches
# pylint: disable=too-many-locals
# pylint: disable=too-many-statements
# validation
if not env.is_cib_live and env.is_corosync_conf_live:
raise LibraryError(
ReportItem.error(
reports.messages.LiveEnvironmentNotConsistent(
[file_type_codes.CIB],
[file_type_codes.COROSYNC_CONF],
)))
if env.is_cib_live and not env.is_corosync_conf_live:
raise LibraryError(
ReportItem.error(
reports.messages.LiveEnvironmentNotConsistent(
[file_type_codes.COROSYNC_CONF],
[file_type_codes.CIB],
)))
# initialization
runner = env.cmd_runner()
report_processor = env.report_processor
live = env.is_cib_live and env.is_corosync_conf_live
is_sbd_running = False
# load status, cib, corosync.conf
status_text, warning_list = get_cluster_status_text(
runner, hide_inactive_resources, verbose)
corosync_conf = None
# If we are live on a remote node, we have no corosync.conf.
# TODO Use the new file framework so the path is not exposed.
if not live or os.path.exists(settings.corosync_conf_file):
corosync_conf = env.get_corosync_conf()
cib = env.get_cib()
if verbose:
(
ticket_status_text,
ticket_status_stderr,
ticket_status_retval,
) = get_ticket_status_text(runner)
# get extra info if live
if live:
try:
is_sbd_running = is_service_running(runner, get_sbd_service_name())
except LibraryError:
pass
local_services_status = _get_local_services_status(runner)
if verbose and corosync_conf:
node_name_list, node_names_report_list = get_existing_nodes_names(
corosync_conf)
report_processor.report_list(node_names_report_list)
node_reachability = _get_node_reachability(
env.get_node_target_factory(),
env.get_node_communicator(),
report_processor,
node_name_list,
)
# check stonith configuration
warning_list = list(warning_list)
warning_list.extend(_stonith_warnings(cib, is_sbd_running))
# put it all together
if report_processor.has_errors:
raise LibraryError()
cluster_name = (corosync_conf.get_cluster_name() if corosync_conf else
nvpair.get_value("cluster_property_set",
get_crm_config(cib), "cluster-name", ""))
parts = []
parts.append(f"Cluster name: {cluster_name}")
if warning_list:
parts.extend(["", "WARNINGS:"] + warning_list + [""])
parts.append(status_text)
if verbose:
parts.extend(["", "Tickets:"])
if ticket_status_retval != 0:
ticket_warning_parts = [
"WARNING: Unable to get information about tickets"
]
if ticket_status_stderr:
ticket_warning_parts.extend(
indent(ticket_status_stderr.splitlines()))
parts.extend(indent(ticket_warning_parts))
else:
parts.extend(indent(ticket_status_text.splitlines()))
if live:
if verbose and corosync_conf:
parts.extend(["", "PCSD Status:"])
parts.extend(
indent(
_format_node_reachability(node_name_list,
node_reachability)))
parts.extend(["", "Daemon Status:"])
parts.extend(
indent(_format_local_services_status(local_services_status)))
return "\n".join(parts)
def update_scsi_devices(
env: LibraryEnvironment,
stonith_id: str,
set_device_list: Iterable[str],
force_flags: Container[reports.types.ForceCode] = (),
) -> None:
"""
Update scsi fencing devices without restart and affecting other resources.
env -- provides all for communication with externals
stonith_id -- id of stonith resource
set_device_list -- paths to the scsi devices that would be set for stonith
resource
force_flags -- list of flags codes
"""
if not is_getting_resource_digest_supported(env.cmd_runner()):
raise LibraryError(
ReportItem.error(
reports.messages.StonithRestartlessUpdateOfScsiDevicesNotSupported()
)
)
cib = env.get_cib()
if not set_device_list:
env.report_processor.report(
ReportItem.error(
reports.messages.InvalidOptionValue(
"devices", "", None, cannot_be_empty=True
)
)
)
(
stonith_el,
report_list,
) = stonith.validate_stonith_restartless_update(cib, stonith_id)
if env.report_processor.report_list(report_list).has_errors:
raise LibraryError()
# for mypy, this should not happen because exeption would be raised
if stonith_el is None:
raise AssertionError("stonith element is None")
stonith.update_scsi_devices_without_restart(
env.cmd_runner(),
env.get_cluster_state(),
stonith_el,
IdProvider(cib),
set_device_list,
)
# Unfencing
cluster_nodes_names, nodes_report_list = get_existing_nodes_names(
env.get_corosync_conf(),
error_on_missing_name=True,
)
env.report_processor.report_list(nodes_report_list)
(
target_report_list,
cluster_nodes_target_list,
) = env.get_node_target_factory().get_target_list_with_reports(
cluster_nodes_names,
allow_skip=False,
)
env.report_processor.report_list(target_report_list)
if env.report_processor.has_errors:
raise LibraryError()
com_cmd: AllSameDataMixin = GetCorosyncOnlineTargets(
env.report_processor,
skip_offline_targets=reports.codes.SKIP_OFFLINE_NODES in force_flags,
)
com_cmd.set_targets(cluster_nodes_target_list)
online_corosync_target_list = run_and_raise(
env.get_node_communicator(), com_cmd
)
com_cmd = Unfence(env.report_processor, sorted(set_device_list))
com_cmd.set_targets(online_corosync_target_list)
run_and_raise(env.get_node_communicator(), com_cmd)
env.push_cib()
def config_sync(
env: LibraryEnvironment, instance_name=None, skip_offline_nodes=False,
):
"""
Send specified local booth configuration to all nodes in the local cluster.
env
string instance_name -- booth instance name
skip_offline_nodes -- if True offline nodes will be skipped
"""
report_processor = env.report_processor
booth_env = env.get_booth_env(instance_name)
if not env.is_cib_live:
raise LibraryError(
ReportItem.error(
reports.messages.LiveEnvironmentRequired([file_type_codes.CIB])
)
)
cluster_nodes_names, report_list = get_existing_nodes_names(
env.get_corosync_conf()
)
if not cluster_nodes_names:
report_list.append(
ReportItem.error(reports.messages.CorosyncConfigNoNodesDefined())
)
report_processor.report_list(report_list)
try:
booth_conf_data = booth_env.config.read_raw()
booth_conf = booth_env.config.raw_to_facade(booth_conf_data)
if isinstance(booth_env.config.raw_file, GhostFile):
authfile_data = booth_env.key.read_raw()
authfile_path = booth_conf.get_authfile()
authfile_name = (
os.path.basename(authfile_path) if authfile_path else None
)
else:
(
authfile_name,
authfile_data,
authfile_report_list,
) = config_files.get_authfile_name_and_data(booth_conf)
report_processor.report_list(authfile_report_list)
except RawFileError as e:
report_processor.report(raw_file_error_report(e))
except ParserErrorException as e:
report_processor.report_list(
booth_env.config.parser_exception_to_report_list(e)
)
if report_processor.has_errors:
raise LibraryError()
com_cmd = BoothSendConfig(
env.report_processor,
booth_env.instance_name,
booth_conf_data,
authfile=authfile_name,
authfile_data=authfile_data,
skip_offline_targets=skip_offline_nodes,
)
com_cmd.set_targets(
env.get_node_target_factory().get_target_list(
cluster_nodes_names, skip_non_existing=skip_offline_nodes,
)
)
run_and_raise(env.get_node_communicator(), com_cmd)
def synchronize_ssl_certificate(env: LibraryEnvironment, skip_offline=False):
"""
Send the local pcsd SSL cert and key to all full nodes in the local cluster.
Consider the pcs Web UI is accessed via an IP running as a resource in the
cluster. When the IP is moved, the user's browser connects to the new node
and we want it to get the same certificate to make the transition a
seamless experience (otherwise the browser display a warning that the
certificate has changed).
Using pcsd Web UI on remote and guest nodes is not supported (pcs/pcsd
depends on the corosanc.conf file being present on the local node) so we
send the cert only to corossync (== full stack) nodes.
"""
report_processor = env.report_processor
target_factory = env.get_node_target_factory()
cluster_nodes_names, report_list = get_existing_nodes_names(
env.get_corosync_conf())
if not cluster_nodes_names:
report_list.append(
ReportItem.error(reports.messages.CorosyncConfigNoNodesDefined()))
report_processor.report_list(report_list)
try:
with open(settings.pcsd_cert_location, "r") as file:
ssl_cert = file.read()
except EnvironmentError as e:
report_processor.report(
ReportItem.error(
reports.messages.FileIoError(
file_type_codes.PCSD_SSL_CERT,
RawFileError.ACTION_READ,
format_environment_error(e),
file_path=settings.pcsd_cert_location,
)))
try:
with open(settings.pcsd_key_location, "r") as file:
ssl_key = file.read()
except EnvironmentError as e:
report_processor.report(
ReportItem.error(
reports.messages.FileIoError(
file_type_codes.PCSD_SSL_KEY,
RawFileError.ACTION_READ,
format_environment_error(e),
file_path=settings.pcsd_key_location,
)))
(
target_report_list,
target_list,
) = target_factory.get_target_list_with_reports(
cluster_nodes_names, skip_non_existing=skip_offline)
report_processor.report_list(target_report_list)
if report_processor.has_errors:
raise LibraryError()
env.report_processor.report(
ReportItem.info(
reports.messages.PcsdSslCertAndKeyDistributionStarted(
sorted([target.label for target in target_list]))))
com_cmd = SendPcsdSslCertAndKey(env.report_processor, ssl_cert, ssl_key)
com_cmd.set_targets(target_list)
run_and_raise(env.get_node_communicator(), com_cmd)
def location_add(lib, argv, modifiers, skip_score_and_node_check=False):
"""
Options:
* --force - allow unknown options, allow constraint for any resource type
* -f - CIB file
"""
del lib
modifiers.ensure_only_supported("--force", "-f")
if len(argv) < 4:
raise CmdLineInputError()
constraint_id = argv.pop(0)
rsc_type, rsc_value = parse_args.parse_typed_arg(
argv.pop(0), [RESOURCE_TYPE_RESOURCE, RESOURCE_TYPE_REGEXP],
RESOURCE_TYPE_RESOURCE)
node = argv.pop(0)
score = argv.pop(0)
options = []
# For now we only allow setting resource-discovery
if argv:
for arg in argv:
if '=' in arg:
options.append(arg.split('=', 1))
else:
raise CmdLineInputError(f"bad option '{arg}'")
if (options[-1][0] != "resource-discovery"
and not modifiers.get("--force")):
utils.err("bad option '%s', use --force to override" %
options[-1][0])
# Verify that specified node exists in the cluster and score is valid
if not skip_score_and_node_check:
if modifiers.is_specified("-f") or modifiers.get("--force"):
warn(LOCATION_NODE_VALIDATION_SKIP_MSG)
else:
lib_env = utils.get_lib_env()
existing_nodes, report_list = get_existing_nodes_names(
corosync_conf=lib_env.get_corosync_conf(),
cib=lib_env.get_cib(),
)
report_list = _verify_node_name(node, existing_nodes)
if report_list:
process_library_reports(report_list)
_verify_score(score)
id_valid, id_error = utils.validate_xml_id(constraint_id, 'constraint id')
if not id_valid:
utils.err(id_error)
required_version = None
if [x for x in options if x[0] == "resource-discovery"]:
required_version = 2, 2, 0
if rsc_type == RESOURCE_TYPE_REGEXP:
required_version = 2, 6, 0
if required_version:
dom = utils.cluster_upgrade_to_version(required_version)
else:
dom = utils.get_cib_dom()
if rsc_type == RESOURCE_TYPE_RESOURCE:
rsc_valid, rsc_error, dummy_correct_id = (
utils.validate_constraint_resource(dom, rsc_value))
if not rsc_valid:
utils.err(rsc_error)
# Verify current constraint doesn't already exist
# If it does we replace it with the new constraint
dummy_dom, constraintsElement = getCurrentConstraints(dom)
elementsToRemove = []
# If the id matches, or the rsc & node match, then we replace/remove
for rsc_loc in constraintsElement.getElementsByTagName('rsc_location'):
# pylint: disable=too-many-boolean-expressions
if (rsc_loc.getAttribute("id") == constraint_id
or (rsc_loc.getAttribute("node") == node and
((RESOURCE_TYPE_RESOURCE == rsc_type
and rsc_loc.getAttribute("rsc") == rsc_value) or
(RESOURCE_TYPE_REGEXP == rsc_type
and rsc_loc.getAttribute("rsc-pattern") == rsc_value)))):
elementsToRemove.append(rsc_loc)
for etr in elementsToRemove:
constraintsElement.removeChild(etr)
element = dom.createElement("rsc_location")
element.setAttribute("id", constraint_id)
if rsc_type == RESOURCE_TYPE_RESOURCE:
element.setAttribute("rsc", rsc_value)
elif rsc_type == RESOURCE_TYPE_REGEXP:
element.setAttribute("rsc-pattern", rsc_value)
element.setAttribute("node", node)
element.setAttribute("score", score)
for option in options:
element.setAttribute(option[0], option[1])
constraintsElement.appendChild(element)
utils.replace_cib_configuration(dom)
def enable_sbd(
lib_env, default_watchdog, watchdog_dict, sbd_options,
default_device_list=None, node_device_dict=None, allow_unknown_opts=False,
ignore_offline_nodes=False, no_watchdog_validation=False,
allow_invalid_option_values=False,
):
# pylint: disable=too-many-arguments, too-many-locals
"""
Enable SBD on all nodes in cluster.
lib_env -- LibraryEnvironment
default_watchdog -- watchdog for nodes which are not specified in
watchdog_dict. Uses default value from settings if None.
watchdog_dict -- dictionary with node names as keys and watchdog path
as value
sbd_options -- dictionary in format: <SBD config option>: <value>
default_device_list -- list of devices for all nodes
node_device_dict -- dictionary with node names as keys and list of devices
as value
allow_unknown_opts -- if True, accept also unknown options.
ignore_offline_nodes -- if True, omit offline nodes
no_watchdog_validation -- it True, do not validate existance of a watchdog
on the nodes
allow_invalid_option_values -- if True, invalid values of some options will
be treated as warning instead of errors
"""
using_devices = not (
default_device_list is None and node_device_dict is None
)
if default_device_list is None:
default_device_list = []
if node_device_dict is None:
node_device_dict = {}
if not default_watchdog:
default_watchdog = settings.sbd_watchdog_default
sbd_options = {opt.upper(): val for opt, val in sbd_options.items()}
corosync_conf = lib_env.get_corosync_conf()
node_list, get_nodes_report_list = get_existing_nodes_names(corosync_conf)
if not node_list:
get_nodes_report_list.append(reports.corosync_config_no_nodes_defined())
target_list = lib_env.get_node_target_factory().get_target_list(
node_list, skip_non_existing=ignore_offline_nodes,
)
full_watchdog_dict = _get_full_target_dict(
target_list, watchdog_dict, default_watchdog
)
full_device_dict = _get_full_target_dict(
target_list, node_device_dict, default_device_list
)
lib_env.report_processor.process_list(
get_nodes_report_list
+
[
reports.node_not_found(node)
for node in (
set(list(watchdog_dict.keys()) + list(node_device_dict.keys()))
-
set(node_list)
)
]
+
_validate_watchdog_dict(full_watchdog_dict)
+
(sbd.validate_nodes_devices(full_device_dict) if using_devices else [])
+
_validate_sbd_options(
sbd_options, allow_unknown_opts, allow_invalid_option_values
)
)
com_cmd = GetOnlineTargets(
lib_env.report_processor, ignore_offline_targets=ignore_offline_nodes,
)
com_cmd.set_targets(target_list)
online_targets = run_and_raise(lib_env.get_node_communicator(), com_cmd)
# check if SBD can be enabled
if no_watchdog_validation:
lib_env.report_processor.report(
reports.sbd_watchdog_validation_inactive()
)
com_cmd = CheckSbd(lib_env.report_processor)
for target in online_targets:
com_cmd.add_request(
target,
(
# Do not send watchdog if validation is turned off. Listing of
# available watchdogs in pcsd may restart the machine in some
# corner cases.
"" if no_watchdog_validation
else full_watchdog_dict[target.label]
),
full_device_dict[target.label] if using_devices else [],
)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# enable ATB if needed
if not using_devices:
if sbd.atb_has_to_be_enabled_pre_enable_check(corosync_conf):
lib_env.report_processor.process(
reports.corosync_quorum_atb_will_be_enabled_due_to_sbd()
)
corosync_conf.set_quorum_options({"auto_tie_breaker": "1"})
lib_env.push_corosync_conf(corosync_conf, ignore_offline_nodes)
# distribute SBD configuration
config = sbd.get_default_sbd_config()
config.update(sbd_options)
com_cmd = SetSbdConfig(lib_env.report_processor)
for target in online_targets:
com_cmd.add_request(
target,
sbd.create_sbd_config(
config,
target.label,
full_watchdog_dict[target.label],
full_device_dict[target.label]
)
)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# remove cluster prop 'stonith_watchdog_timeout'
com_cmd = RemoveStonithWatchdogTimeout(lib_env.report_processor)
com_cmd.set_targets(online_targets)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# enable SBD service an all nodes
com_cmd = EnableSbdService(lib_env.report_processor)
com_cmd.set_targets(online_targets)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
lib_env.report_processor.process(
reports.cluster_restart_required_to_apply_changes()
)
def add_device(
lib_env: LibraryEnvironment,
model,
model_options,
generic_options,
heuristics_options,
force_model=False,
force_options=False,
skip_offline_nodes=False,
):
# pylint: disable=too-many-locals
"""
Add a quorum device to a cluster, distribute and reload configs if live
string model -- quorum device model
dict model_options -- model specific options
dict generic_options -- generic quorum device options
dict heuristics_options -- heuristics options
bool force_model -- continue even if the model is not valid
bool force_options -- continue even if options are not valid
bool skip_offline_nodes -- continue even if not all nodes are accessible
"""
cfg = lib_env.get_corosync_conf()
if cfg.has_quorum_device():
raise LibraryError(
ReportItem.error(reports.messages.QdeviceAlreadyDefined()))
report_processor = lib_env.report_processor
report_processor.report_list(
corosync_conf_validators.add_quorum_device(
model,
model_options,
generic_options,
heuristics_options,
[node.nodeid for node in cfg.get_nodes()],
force_model=force_model,
force_options=force_options,
))
if lib_env.is_corosync_conf_live:
cluster_nodes_names, report_list = get_existing_nodes_names(
cfg,
# Pcs is unable to communicate with nodes missing names. It cannot
# send new corosync.conf to them. That might break the cluster.
# Hence we error out.
error_on_missing_name=True,
)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
cfg.add_quorum_device(
model,
model_options,
generic_options,
heuristics_options,
)
if cfg.is_quorum_device_heuristics_enabled_with_no_exec():
lib_env.report_processor.report(
ReportItem.warning(
reports.messages.CorosyncQuorumHeuristicsEnabledWithNoExec()))
# First setup certificates for qdevice, then send corosync.conf to nodes.
# If anything fails, nodes will not have corosync.conf with qdevice in it,
# so there is no effect on the cluster.
if lib_env.is_corosync_conf_live:
target_factory = lib_env.get_node_target_factory()
target_list = target_factory.get_target_list(
cluster_nodes_names,
skip_non_existing=skip_offline_nodes,
)
# Do model specific configuration.
# If the model is not known to pcs and was forced, do not configure
# anything else than corosync.conf, as we do not know what to do
# anyway.
if model == "net":
qdevice_net.set_up_client_certificates(
lib_env.cmd_runner(),
lib_env.report_processor,
lib_env.communicator_factory,
# We are sure the "host" key is there, it has been validated
# above.
target_factory.get_target_from_hostname(model_options["host"]),
cfg.get_cluster_name(),
target_list,
skip_offline_nodes,
)
lib_env.report_processor.report(
ReportItem.info(
reports.messages.ServiceActionStarted(
reports.const.SERVICE_ACTION_ENABLE, "corosync-qdevice")))
com_cmd = qdevice_com.Enable(lib_env.report_processor,
skip_offline_nodes)
com_cmd.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# everything set up, it's safe to tell the nodes to use qdevice
lib_env.push_corosync_conf(cfg, skip_offline_nodes)
# Now, when corosync.conf has been reloaded, we can start qdevice service.
if lib_env.is_corosync_conf_live:
lib_env.report_processor.report(
ReportItem.info(
reports.messages.ServiceActionStarted(
reports.const.SERVICE_ACTION_START, "corosync-qdevice")))
com_cmd_start = qdevice_com.Start(lib_env.report_processor,
skip_offline_nodes)
com_cmd_start.set_targets(target_list)
run_and_raise(lib_env.get_node_communicator(), com_cmd_start)
def config_restore_remote(infile_name, infile_obj):
"""
Commandline options:
* --request-timeout - timeout for HTTP requests
"""
extracted = {
"version.txt": "",
"corosync.conf": "",
}
try:
tarball = tarfile.open(infile_name, "r|*", infile_obj)
while True:
# next(tarball) does not work in python2.6
tar_member_info = tarball.next()
if tar_member_info is None:
break
if tar_member_info.name in extracted:
tar_member = tarball.extractfile(tar_member_info)
extracted[tar_member_info.name] = tar_member.read()
tar_member.close()
tarball.close()
except (tarfile.TarError, EnvironmentError) as e:
utils.err("unable to read the tarball: %s" % e)
config_backup_check_version(extracted["version.txt"])
node_list, report_list = get_existing_nodes_names(
utils.get_corosync_conf_facade(
conf_text=extracted["corosync.conf"].decode("utf-8")))
if report_list:
process_library_reports(report_list)
if not node_list:
utils.err("no nodes found in the tarball")
err_msgs = []
for node in node_list:
try:
retval, output = utils.checkStatus(node)
if retval != 0:
err_msgs.append(output)
continue
_status = json.loads(output)
if (_status["corosync"] or _status["pacemaker"] or
# not supported by older pcsd, do not fail if not present
_status.get("pacemaker_remote", False)):
err_msgs.append(
"Cluster is currently running on node %s. You need to stop "
"the cluster in order to restore the configuration." %
node)
continue
except (ValueError, NameError, LookupError):
err_msgs.append("unable to determine status of the node %s" % node)
if err_msgs:
for msg in err_msgs:
utils.err(msg, False)
sys.exit(1)
# Temporarily disable config files syncing thread in pcsd so it will not
# rewrite restored files. 10 minutes should be enough time to restore.
# If node returns HTTP 404 it does not support config syncing at all.
for node in node_list:
retval, output = utils.pauseConfigSyncing(node, 10 * 60)
if not (retval == 0 or "(HTTP error: 404)" in output):
utils.err(output)
if infile_obj:
infile_obj.seek(0)
tarball_data = infile_obj.read()
else:
with open(infile_name, "rb") as tarball:
tarball_data = tarball.read()
error_list = []
for node in node_list:
retval, error = utils.restoreConfig(node, tarball_data)
if retval != 0:
error_list.append(error)
if error_list:
utils.err("unable to restore all nodes\n" + "\n".join(error_list))
def config_restore_remote(infile_name, infile_obj):
"""
Commandline options:
* --request-timeout - timeout for HTTP requests
"""
extracted = {
"version.txt": "",
"corosync.conf": "",
}
try:
tarball = tarfile.open(infile_name, "r|*", infile_obj)
while True:
# next(tarball) does not work in python2.6
tar_member_info = tarball.next()
if tar_member_info is None:
break
if tar_member_info.name in extracted:
tar_member = tarball.extractfile(tar_member_info)
extracted[tar_member_info.name] = tar_member.read()
tar_member.close()
tarball.close()
except (tarfile.TarError, EnvironmentError) as e:
utils.err("unable to read the tarball: %s" % e)
config_backup_check_version(extracted["version.txt"])
node_list, report_list = get_existing_nodes_names(
utils.get_corosync_conf_facade(
conf_text=extracted["corosync.conf"].decode("utf-8")
)
)
if report_list:
utils.process_library_reports(report_list)
if not node_list:
utils.err("no nodes found in the tarball")
err_msgs = []
for node in node_list:
try:
retval, output = utils.checkStatus(node)
if retval != 0:
err_msgs.append(output)
continue
_status = json.loads(output)
if (
_status["corosync"]
or
_status["pacemaker"]
or
# not supported by older pcsd, do not fail if not present
_status.get("pacemaker_remote", False)
):
err_msgs.append(
"Cluster is currently running on node %s. You need to stop "
"the cluster in order to restore the configuration."
% node
)
continue
except (ValueError, NameError, LookupError):
err_msgs.append("unable to determine status of the node %s" % node)
if err_msgs:
for msg in err_msgs:
utils.err(msg, False)
sys.exit(1)
# Temporarily disable config files syncing thread in pcsd so it will not
# rewrite restored files. 10 minutes should be enough time to restore.
# If node returns HTTP 404 it does not support config syncing at all.
for node in node_list:
retval, output = utils.pauseConfigSyncing(node, 10 * 60)
if not (retval == 0 or "(HTTP error: 404)" in output):
utils.err(output)
if infile_obj:
infile_obj.seek(0)
tarball_data = infile_obj.read()
else:
with open(infile_name, "rb") as tarball:
tarball_data = tarball.read()
error_list = []
for node in node_list:
retval, error = utils.restoreConfig(node, tarball_data)
if retval != 0:
error_list.append(error)
if error_list:
utils.err("unable to restore all nodes\n" + "\n".join(error_list))
def set_recovery_site(env: LibraryEnvironment, node_name: str) -> None:
"""
Set up disaster recovery with the local cluster being the primary site
env
node_name -- a known host from the recovery site
"""
if env.ghost_file_codes:
raise LibraryError(
ReportItem.error(
reports.messages.LiveEnvironmentRequired(
env.ghost_file_codes)))
report_processor = env.report_processor
dr_env = env.get_dr_env()
if dr_env.config.raw_file.exists():
report_processor.report(
ReportItem.error(reports.messages.DrConfigAlreadyExist()))
target_factory = env.get_node_target_factory()
local_nodes, report_list = get_existing_nodes_names(
env.get_corosync_conf(), error_on_missing_name=True)
report_processor.report_list(report_list)
if node_name in local_nodes:
report_processor.report(
ReportItem.error(reports.messages.NodeInLocalCluster(node_name)))
report_list, local_targets = target_factory.get_target_list_with_reports(
local_nodes, allow_skip=False, report_none_host_found=False)
report_processor.report_list(report_list)
report_list, remote_targets = target_factory.get_target_list_with_reports(
[node_name], allow_skip=False, report_none_host_found=False)
report_processor.report_list(report_list)
if report_processor.has_errors:
raise LibraryError()
com_cmd = GetCorosyncConf(env.report_processor)
com_cmd.set_targets(remote_targets)
remote_cluster_nodes, report_list = get_existing_nodes_names(
CorosyncConfigFacade.from_string(
run_and_raise(env.get_node_communicator(), com_cmd)),
error_on_missing_name=True,
)
if report_processor.report_list(report_list).has_errors:
raise LibraryError()
# ensure we have tokens for all nodes of remote cluster
report_list, remote_targets = target_factory.get_target_list_with_reports(
remote_cluster_nodes, allow_skip=False, report_none_host_found=False)
if report_processor.report_list(report_list).has_errors:
raise LibraryError()
dr_config_exporter = get_file_toolbox(
file_type_codes.PCS_DR_CONFIG).exporter
# create dr config for remote cluster
remote_dr_cfg = dr_env.create_facade(DrRole.RECOVERY)
remote_dr_cfg.add_site(DrRole.PRIMARY, local_nodes)
# send config to all node of remote cluster
distribute_file_cmd = DistributeFilesWithoutForces(
env.report_processor,
node_communication_format.pcs_dr_config_file(
dr_config_exporter.export(remote_dr_cfg.config)),
)
distribute_file_cmd.set_targets(remote_targets)
run_and_raise(env.get_node_communicator(), distribute_file_cmd)
# create new dr config, with local cluster as primary site
local_dr_cfg = dr_env.create_facade(DrRole.PRIMARY)
local_dr_cfg.add_site(DrRole.RECOVERY, remote_cluster_nodes)
distribute_file_cmd = DistributeFilesWithoutForces(
env.report_processor,
node_communication_format.pcs_dr_config_file(
dr_config_exporter.export(local_dr_cfg.config)),
)
distribute_file_cmd.set_targets(local_targets)
run_and_raise(env.get_node_communicator(), distribute_file_cmd)
def enable_sbd(
lib_env,
default_watchdog,
watchdog_dict,
sbd_options,
default_device_list=None,
node_device_dict=None,
allow_unknown_opts=False,
ignore_offline_nodes=False,
no_watchdog_validation=False,
allow_invalid_option_values=False,
):
# pylint: disable=too-many-arguments, too-many-locals
"""
Enable SBD on all nodes in cluster.
lib_env -- LibraryEnvironment
default_watchdog -- watchdog for nodes which are not specified in
watchdog_dict. Uses default value from settings if None.
watchdog_dict -- dictionary with node names as keys and watchdog path
as value
sbd_options -- dictionary in format: <SBD config option>: <value>
default_device_list -- list of devices for all nodes
node_device_dict -- dictionary with node names as keys and list of devices
as value
allow_unknown_opts -- if True, accept also unknown options.
ignore_offline_nodes -- if True, omit offline nodes
no_watchdog_validation -- it True, do not validate existance of a watchdog
on the nodes
allow_invalid_option_values -- if True, invalid values of some options will
be treated as warning instead of errors
"""
using_devices = not (default_device_list is None
and node_device_dict is None)
if default_device_list is None:
default_device_list = []
if node_device_dict is None:
node_device_dict = {}
if not default_watchdog:
default_watchdog = settings.sbd_watchdog_default
sbd_options = {opt.upper(): val for opt, val in sbd_options.items()}
corosync_conf = lib_env.get_corosync_conf()
node_list, get_nodes_report_list = get_existing_nodes_names(corosync_conf)
if not node_list:
get_nodes_report_list.append(
ReportItem.error(reports.messages.CorosyncConfigNoNodesDefined()))
target_list = lib_env.get_node_target_factory().get_target_list(
node_list,
skip_non_existing=ignore_offline_nodes,
)
full_watchdog_dict = _get_full_target_dict(target_list, watchdog_dict,
default_watchdog)
full_device_dict = _get_full_target_dict(target_list, node_device_dict,
default_device_list)
if lib_env.report_processor.report_list(
get_nodes_report_list +
[
ReportItem.error(reports.messages.NodeNotFound(node)) for node in
(set(list(watchdog_dict.keys()) + list(node_device_dict.keys())) -
set(node_list))
] + _validate_watchdog_dict(full_watchdog_dict) +
(sbd.validate_nodes_devices(full_device_dict) if using_devices else []
) + _validate_sbd_options(sbd_options, allow_unknown_opts,
allow_invalid_option_values)).has_errors:
raise LibraryError()
com_cmd = GetOnlineTargets(
lib_env.report_processor,
ignore_offline_targets=ignore_offline_nodes,
)
com_cmd.set_targets(target_list)
online_targets = run_and_raise(lib_env.get_node_communicator(), com_cmd)
# check if SBD can be enabled
if no_watchdog_validation:
lib_env.report_processor.report(
ReportItem.warning(
reports.messages.SbdWatchdogValidationInactive()))
com_cmd = CheckSbd(lib_env.report_processor)
for target in online_targets:
com_cmd.add_request(
target,
(
# Do not send watchdog if validation is turned off. Listing of
# available watchdogs in pcsd may restart the machine in some
# corner cases.
"" if no_watchdog_validation else
full_watchdog_dict[target.label]),
full_device_dict[target.label] if using_devices else [],
)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# enable ATB if needed
if not using_devices:
if sbd.atb_has_to_be_enabled_pre_enable_check(corosync_conf):
lib_env.report_processor.report(
ReportItem.warning(
reports.messages.CorosyncQuorumAtbWillBeEnabledDueToSbd()))
corosync_conf.set_quorum_options({"auto_tie_breaker": "1"})
lib_env.push_corosync_conf(corosync_conf, ignore_offline_nodes)
# distribute SBD configuration
config = sbd.get_default_sbd_config()
config.update(sbd_options)
com_cmd = SetSbdConfig(lib_env.report_processor)
for target in online_targets:
com_cmd.add_request(
target,
sbd.create_sbd_config(
config,
target.label,
full_watchdog_dict[target.label],
full_device_dict[target.label],
),
)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# remove cluster prop 'stonith_watchdog_timeout'
com_cmd = RemoveStonithWatchdogTimeout(lib_env.report_processor)
com_cmd.set_targets(online_targets)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
# enable SBD service an all nodes
com_cmd = EnableSbdService(lib_env.report_processor)
com_cmd.set_targets(online_targets)
run_and_raise(lib_env.get_node_communicator(), com_cmd)
lib_env.report_processor.report(
ReportItem.warning(
reports.messages.ClusterRestartRequiredToApplyChanges()))
