def _find_host_candidates(
broker: ParsedSectionsBroker,
preliminary_candidates: List[checking_classes.CheckPlugin],
parsed_sections_of_interest: Set[ParsedSectionName],
) -> Set[CheckPluginName]:
available_parsed_sections = broker.filter_available(
parsed_sections_of_interest,
SourceType.HOST,
)
return {
plugin.name
for plugin in preliminary_candidates
# *filter out* all names of management only check plugins
if not plugin.name.is_management_name() and any(
section in available_parsed_sections
for section in plugin.sections)
}
def _discover_host_labels_for_source_type(
*,
host_key: HostKey,
parsed_sections_broker: ParsedSectionsBroker,
on_error: OnError,
) -> Mapping[str, HostLabel]:
host_labels = {}
try:
parsed_results = parsed_sections_broker.all_parsing_results(host_key)
console.vverbose("Trying host label discovery with: %s\n" %
", ".join(str(r.section.name) for r in parsed_results))
for (section_data, _cache_info), section_plugin in parsed_results:
kwargs = {'section': section_data}
host_label_params = config.get_host_label_parameters(host_key.hostname, section_plugin)
if host_label_params is not None:
kwargs["params"] = host_label_params
try:
for label in section_plugin.host_label_function(**kwargs):
console.vverbose(f" {label.name}: {label.value} ({section_plugin.name})\n")
host_labels[label.name] = HostLabel(
label.name,
label.value,
section_plugin.name,
)
except (KeyboardInterrupt, MKTimeout):
raise
except Exception as exc:
if on_error is OnError.RAISE:
raise
if on_error is OnError.WARN:
console.error(
f"Host label discovery of '{section_plugin.name}' failed: {exc}\n")
except KeyboardInterrupt:
raise MKGeneralException("Interrupted by Ctrl-C.")
return host_labels
def test_multiple_sources_from_different_hosts(self, hostname, ipaddress, mode, config_cache, host_config):
sources = [
ProgramSource.ds(hostname + "0", ipaddress, mode=mode, template=""),
TCPSource(hostname + "1", ipaddress, mode=mode),
TCPSource(hostname + "2", ipaddress, mode=mode),
]
nodes = make_nodes(
config_cache,
host_config,
ipaddress,
mode=mode,
sources=sources,
)
broker = ParsedSectionsBroker(_collect_host_sections(
nodes=nodes,
file_cache_max_age=0,
fetcher_messages=[
FetcherMessage.from_raw_data(
result.OK(source.default_raw_data),
Snapshot.null(),
source.fetcher_type,
)
for _h, _i, sources in nodes for source in sources
],
selected_sections=NO_SELECTION,
)[0])
assert len(broker) == 1
key = HostKey(hostname, ipaddress, SourceType.HOST)
assert key in broker
section = broker[key]
assert len(section.sections) == len(sources)
for source in sources:
# yapf: disable
assert (
section.sections[SectionName("section_name_%s" % source.hostname)]
== [["section_content"]])
def test_get_section_content(hostname, host_entries, cluster_node_keys, expected_result):
parsed_sections_broker = ParsedSectionsBroker({
HostKey(nodename, "127.0.0.1",
SourceType.HOST): SectionsParser(host_sections=AgentHostSections(
sections={SectionName("section_plugin_name"): node_section_content}))
for nodename, node_section_content in host_entries
})
mhs = _MultiHostSections(parsed_sections_broker)
section_content = mhs.get_section_content(
HostKey(hostname, "127.0.0.1", SourceType.HOST),
HOST_ONLY,
"section_plugin_name",
False,
cluster_node_keys=cluster_node_keys,
check_legacy_info={}, # only for parse_function lookup, not needed in this test
)
assert expected_result == section_content
section_content = mhs.get_section_content(
HostKey(hostname, "127.0.0.1", SourceType.HOST),
HOST_PRECEDENCE,
"section_plugin_name",
False,
cluster_node_keys=cluster_node_keys,
check_legacy_info={}, # only for parse_function lookup, not needed in this test
)
assert expected_result == section_content
section_content = mhs.get_section_content(
HostKey(hostname, "127.0.0.1", SourceType.MANAGEMENT),
MGMT_ONLY,
"section_plugin_name",
False,
cluster_node_keys=None if cluster_node_keys is None else
[HostKey(hn, ip, SourceType.MANAGEMENT) for (hn, ip, _st) in cluster_node_keys],
check_legacy_info={}, # only for parse_function lookup, not needed in this test
)
assert section_content is None
def _find_mgmt_candidates(
broker: ParsedSectionsBroker,
preliminary_candidates: List[checking_classes.CheckPlugin],
parsed_sections_of_interest: Set[ParsedSectionName],
) -> Set[CheckPluginName]:
available_parsed_sections = {
s.parsed_section_name
for s in broker.determine_applicable_sections(
parsed_sections_of_interest,
SourceType.MANAGEMENT,
)
}
return {
# *create* all management only names of the plugins
plugin.name.create_management_name()
for plugin in preliminary_candidates
if any(section in available_parsed_sections
for section in plugin.sections)
}
def test_one_snmp_source(self, hostname, ipaddress, mode, config_cache,
host_config):
broker = ParsedSectionsBroker(
_collect_host_sections(
nodes=make_nodes(
config_cache,
host_config,
ipaddress,
mode=mode,
sources=[
SNMPSource.snmp(
hostname,
ipaddress,
mode=mode,
selected_sections=NO_SELECTION,
force_cache_refresh=False,
on_scan_error="raise",
),
],
),
file_cache_max_age=0,
fetcher_messages=[
FetcherMessage.from_raw_data(
result.OK({}),
Snapshot.null(),
FetcherType.SNMP,
),
],
selected_sections=NO_SELECTION,
)[0])
assert len(broker) == 1
key = HostKey(hostname, ipaddress, SourceType.HOST)
assert key in broker
section = broker[key]
assert len(section.sections) == 1
assert section.sections[SectionName("section_name_%s" %
hostname)] == [["section_content"]]
def test_get_section_cluster_kwargs(required_sections, expected_result):
node1_sections = AgentHostSections(
sections={
SectionName("one"): NODE_1,
SectionName("two"): NODE_1,
SectionName("three"): NODE_1
})
node2_sections = AgentHostSections(sections={
SectionName("two"): NODE_2,
SectionName("three"): NODE_2,
})
parsed_sections_broker = ParsedSectionsBroker({
HostKey("node1", "127.0.0.1", SourceType.HOST): (
ParsedSectionsResolver(section_plugins=[
SECTION_ONE, SECTION_TWO, SECTION_THREE, SECTION_FOUR
], ),
SectionsParser(host_sections=node1_sections),
),
HostKey("node2", "127.0.0.1", SourceType.HOST): (
ParsedSectionsResolver(section_plugins=[
SECTION_ONE, SECTION_TWO, SECTION_THREE, SECTION_FOUR
], ),
SectionsParser(host_sections=node2_sections),
),
})
kwargs = get_section_cluster_kwargs(
parsed_sections_broker,
[
HostKey("node1", "127.0.0.1", SourceType.HOST),
HostKey("node2", "127.0.0.1", SourceType.HOST),
],
[ParsedSectionName(n) for n in required_sections],
)
assert expected_result == kwargs
def test_no_sources(self, cluster, nodes, config_cache, host_config, mode):
made_nodes = make_nodes(
config_cache,
host_config,
None,
mode=mode,
sources=(),
)
broker = ParsedSectionsBroker(_collect_host_sections(
nodes=made_nodes,
file_cache_max_age=0,
fetcher_messages=[
# We do not pass sources explicitly but still append Piggyback.
FetcherMessage.from_raw_data(
result.OK(AgentRawData(b"")),
Snapshot.null(),
FetcherType.PIGGYBACK,
) for _n in made_nodes
],
selected_sections=NO_SELECTION,
)[0])
assert len(broker) == len(nodes)
key_clu = HostKey(cluster, None, SourceType.HOST)
assert key_clu not in broker
for hostname, addr in nodes.items():
key = HostKey(hostname, addr, SourceType.HOST)
assert key in broker
section = broker[key]
# yapf: disable
assert (section.sections[SectionName("section_name_%s" % hostname)]
== [["section_content_%s" % hostname]])
assert not section.cache_info
assert not section.piggybacked_raw_data
def test__find_candidates():
broker = ParsedSectionsBroker({
# we just care about the keys here, content set to arbitrary values that can be parsed.
# section names are chosen arbitrarily.
HostKey("test_node", "1.2.3.4", SourceType.HOST):
SectionsParser(
host_sections=AgentHostSections({
SectionName("kernel"): [], # host only
SectionName("uptime"): [['123']], # host & mgmt
}), ),
HostKey("test_node", "1.2.3.4", SourceType.MANAGEMENT):
SectionsParser(
host_sections=SNMPHostSections({
# host & mgmt:
SectionName("uptime"): [['123']], # type: ignore[dict-item]
# mgmt only:
SectionName("liebert_fans"):
[[['Fan', '67', 'umin']]], # type: ignore[dict-item]
# is already mgmt_ prefixed:
SectionName("mgmt_snmp_info"):
[[['a', 'b', 'c', 'd']]], # type: ignore[dict-item]
}), ),
})
preliminary_candidates = list(
agent_based_register.iter_all_check_plugins())
parsed_sections_of_interest = {
parsed_section_name
for plugin in preliminary_candidates
for parsed_section_name in plugin.sections
}
assert discovery._discovered_services._find_host_candidates(
broker,
preliminary_candidates,
parsed_sections_of_interest,
) == {
CheckPluginName('docker_container_status_uptime'),
CheckPluginName("kernel"),
CheckPluginName('kernel_performance'),
CheckPluginName('kernel_util'),
CheckPluginName("uptime"),
}
assert discovery._discovered_services._find_mgmt_candidates(
broker,
preliminary_candidates,
parsed_sections_of_interest,
) == {
CheckPluginName('mgmt_docker_container_status_uptime'),
CheckPluginName("mgmt_liebert_fans"),
CheckPluginName("mgmt_uptime"),
}
assert discovery._discovered_services._find_candidates(
broker,
run_plugin_names=EVERYTHING,
) == {
CheckPluginName('docker_container_status_uptime'),
CheckPluginName("kernel"),
CheckPluginName('kernel_performance'),
CheckPluginName('kernel_util'),
CheckPluginName('mgmt_docker_container_status_uptime'),
CheckPluginName("mgmt_liebert_fans"),
CheckPluginName("mgmt_uptime"),
CheckPluginName("uptime"),
}
def get_aggregated_result(
parsed_sections_broker: ParsedSectionsBroker,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
service: Service,
plugin: Optional[checking_classes.CheckPlugin],
params_function: Callable[[], Parameters],
*,
value_store_manager: item_state.ValueStoreManager,
) -> AggregatedResult:
"""Run the check function and aggregate the subresults
This function is also called during discovery.
"""
if plugin is None:
return AggregatedResult(
submit=True,
data_received=True,
result=CHECK_NOT_IMPLEMENTED,
cache_info=None,
)
check_function = (plugin.cluster_check_function
if host_config.is_cluster else plugin.check_function)
source_type = (SourceType.MANAGEMENT
if service.check_plugin_name.is_management_name() else SourceType.HOST)
config_cache = config.get_config_cache()
kwargs = {}
try:
kwargs = parsed_sections_broker.get_section_cluster_kwargs(
config_cache.get_clustered_service_node_keys(
host_config.hostname,
source_type,
service.description,
) or [],
plugin.sections,
) if host_config.is_cluster else parsed_sections_broker.get_section_kwargs(
HostKey(host_config.hostname, ipaddress, source_type),
plugin.sections,
)
if not kwargs and not service.check_plugin_name.is_management_name():
# in 1.6 some plugins where discovered for management boards, but with
# the regular host plugins name. In this case retry with the source type
# forced to MANAGEMENT:
kwargs = parsed_sections_broker.get_section_cluster_kwargs(
config_cache.get_clustered_service_node_keys(
host_config.hostname,
SourceType.MANAGEMENT,
service.description,
) or [],
plugin.sections,
) if host_config.is_cluster else parsed_sections_broker.get_section_kwargs(
HostKey(host_config.hostname, ipaddress, SourceType.MANAGEMENT),
plugin.sections,
)
if not kwargs: # no data found
return AggregatedResult(
submit=False,
data_received=False,
result=RECEIVED_NO_DATA,
cache_info=None,
)
if service.item is not None:
kwargs["item"] = service.item
if plugin.check_default_parameters is not None:
kwargs["params"] = params_function()
with plugin_contexts.current_host(host_config.hostname), \
plugin_contexts.current_service(service), \
value_store_manager.namespace(service.id()):
result = _aggregate_results(check_function(**kwargs))
except (item_state.MKCounterWrapped, checking_classes.IgnoreResultsError) as e:
msg = str(e) or "No service summary available"
return AggregatedResult(
submit=False,
data_received=True,
result=(0, msg, []),
cache_info=None,
)
except MKTimeout:
raise
except Exception:
if cmk.utils.debug.enabled():
raise
table = check_table.get_check_table(host_config.hostname, skip_autochecks=True)
result = 3, cmk.base.crash_reporting.create_check_crash_dump(
host_name=host_config.hostname,
service_name=service.description,
plugin_name=service.check_plugin_name,
plugin_kwargs=kwargs,
is_manual=service.id() in table,
), []
return AggregatedResult(
submit=True,
data_received=True,
result=result,
cache_info=parsed_sections_broker.get_cache_info(plugin.sections),
)
def _realhost_scenario(monkeypatch):
hostname = "test-realhost"
ipaddress = "1.2.3.4"
ts = Scenario().add_host(hostname, ipaddress=ipaddress)
ts.set_ruleset("inventory_df_rules", [{
'value': {
'ignore_fs_types': ['tmpfs', 'nfs', 'smbfs', 'cifs', 'iso9660'],
'never_ignore_mountpoints': ['~.*/omd/sites/[^/]+/tmp$']
},
'condition': {
'host_labels': {
'cmk/check_mk_server': 'yes'
}
}
}])
ts.apply(monkeypatch)
def fake_lookup_ip_address(*_a, **_kw):
return ipaddress
monkeypatch.setattr(ip_lookup, "lookup_ip_address", fake_lookup_ip_address)
DiscoveredHostLabelsStore(hostname).save({
'existing_label': {
'plugin_name': 'foo',
'value': 'bar',
},
'another_label': {
'plugin_name': 'labels',
'value': 'true',
}
})
broker = ParsedSectionsBroker({
HostKey(hostname=hostname,
ipaddress=ipaddress,
source_type=SourceType.HOST):
AgentHostSections(
sections={
SectionName("labels"): [
[
'{"cmk/check_mk_server":"yes"}',
],
],
SectionName("df"): [
[
'/dev/sda1',
'vfat',
'523248',
'3668',
'519580',
'1%',
'/boot/test-efi',
],
[
'tmpfs',
'tmpfs',
'8152916',
'244',
'8152672',
'1%',
'/opt/omd/sites/test-heute/tmp',
],
],
})
})
return RealHostScenario(hostname, ipaddress, broker)
def check_discovery(
host_name: HostName,
ipaddress: Optional[HostAddress],
*,
# The next argument *must* remain optional for the DiscoCheckExecutor.
# See Also: `cmk.base.checking.do_check()`.
fetcher_messages: Sequence[FetcherMessage] = (),
) -> Tuple[int, List[str], List[str], List[Tuple]]:
# Note: '--cache' is set in core_cmc, nagios template or even on CL and means:
# 1. use caches as default:
# - Set FileCacheFactory.maybe = True (set max_cachefile_age, else 0)
# - Set FileCacheFactory.use_outdated = True
# 2. Then these settings are used to read cache file or not
config_cache = config.get_config_cache()
host_config = config_cache.get_host_config(host_name)
discovery_parameters = DiscoveryParameters(
on_error="raise",
load_labels=True,
save_labels=False,
only_host_labels=False,
)
params = host_config.discovery_check_parameters
if params is None:
params = host_config.default_discovery_check_parameters()
# In case of keepalive discovery we always have an ipaddress. When called as non keepalive
# ipaddress is always None
if ipaddress is None and not host_config.is_cluster:
ipaddress = config.lookup_ip_address(host_config)
nodes = sources.make_nodes(
config_cache,
host_config,
ipaddress,
Mode.DISCOVERY,
sources.make_sources(host_config, ipaddress, mode=Mode.DISCOVERY),
)
use_caches = cmk.core_helpers.cache.FileCacheFactory.maybe
max_cachefile_age = config.discovery_max_cachefile_age() if use_caches else 0
if not fetcher_messages:
# Note: *Not* calling `fetch_all(sources)` here is probably buggy.
# Also See: `cmk.base.checking.do_check()`
fetcher_messages = list(
sources.fetch_all(
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
))
parsed_sections_broker = ParsedSectionsBroker()
result = sources.update_host_sections(
parsed_sections_broker,
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
fetcher_messages=fetcher_messages,
selected_sections=NO_SELECTION,
)
services, host_label_discovery_result = _get_host_services(
host_config,
ipaddress,
parsed_sections_broker,
discovery_parameters,
)
status, infotexts, long_infotexts, perfdata, need_rediscovery = _aggregate_subresults(
_check_service_lists(host_name, services, params),
_check_host_labels(host_label_discovery_result, params),
_check_data_sources(result),
)
if need_rediscovery:
if host_config.is_cluster and host_config.nodes:
for nodename in host_config.nodes:
_set_rediscovery_flag(nodename)
else:
_set_rediscovery_flag(host_name)
infotexts.append(u"rediscovery scheduled")
return status, infotexts, long_infotexts, perfdata
def test_get_host_sections_cluster(mode, monkeypatch, mocker):
hostname = "testhost"
hosts = {
"host0": "10.0.0.0",
"host1": "10.0.0.1",
"host2": "10.0.0.2",
}
address = "1.2.3.4"
tags = {"agent": "no-agent"}
section_name = SectionName("test_section")
config_cache = make_scenario(hostname, tags).apply(monkeypatch)
host_config = config.HostConfig.make_host_config(hostname)
def fake_lookup_ip_address(host_config, family=None):
return hosts[host_config.hostname]
def check(_, *args, **kwargs):
return result.OK(AgentHostSections(sections={section_name: [[str(section_name)]]}))
monkeypatch.setattr(
config,
"lookup_ip_address",
fake_lookup_ip_address,
)
monkeypatch.setattr(
Source,
"parse",
check,
)
mocker.patch.object(
cmk.utils.piggyback,
"remove_source_status_file",
autospec=True,
)
mocker.patch.object(
cmk.utils.piggyback,
"_store_status_file_of",
autospec=True,
)
# Create a cluster
host_config.nodes = list(hosts.keys())
nodes = make_nodes(
config_cache,
host_config,
address,
mode=mode,
sources=make_sources(host_config, address, mode=mode)
)
broker = ParsedSectionsBroker(_collect_host_sections(
nodes=nodes,
file_cache_max_age=host_config.max_cachefile_age,
fetcher_messages=[
FetcherMessage.from_raw_data(
result.OK(source.default_raw_data),
Snapshot.null(),
source.fetcher_type,
)
for _h, _i, sources in nodes for source in sources
],
selected_sections=NO_SELECTION,
)[0])
assert len(broker) == len(hosts) == 3
cmk.utils.piggyback._store_status_file_of.assert_not_called() # type: ignore[attr-defined]
assert cmk.utils.piggyback.remove_source_status_file.call_count == 3 # type: ignore[attr-defined]
for host, addr in hosts.items():
remove_source_status_file = cmk.utils.piggyback.remove_source_status_file
remove_source_status_file.assert_any_call(host) # type: ignore[attr-defined]
key = HostKey(host, addr, SourceType.HOST)
assert key in broker
section = broker[key]
assert len(section.sections) == 1
assert next(iter(section.sections)) == section_name
assert not section.cache_info
assert not section.piggybacked_raw_data
def get_aggregated_result(
parsed_sections_broker: ParsedSectionsBroker,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
service: Service,
plugin: Optional[checking_classes.CheckPlugin],
# missleading. These are prams that *may* be *partially* time specific
timespecific_parameters: LegacyCheckParameters,
*,
value_store_manager: value_store.ValueStoreManager,
persist_value_store_changes: bool,
) -> AggregatedResult:
"""Run the check function and aggregate the subresults
This function is also called during discovery.
"""
if plugin is None:
return AggregatedResult(
submit=True,
data_received=True,
result=CHECK_NOT_IMPLEMENTED,
cache_info=None,
)
config_cache = config.get_config_cache()
check_function = (
_cluster_modes.get_cluster_check_function(
*config_cache.get_clustered_service_configuration(
host_config.hostname,
service.description,
),
plugin=plugin,
service_id=service.id(),
persist_value_store_changes=persist_value_store_changes,
)
if host_config.is_cluster
else plugin.check_function
)
source_type = (
SourceType.MANAGEMENT if service.check_plugin_name.is_management_name() else SourceType.HOST
)
try:
kwargs = (
get_section_cluster_kwargs(
parsed_sections_broker,
config_cache.get_clustered_service_node_keys(
host_config.hostname,
source_type,
service.description,
)
or [],
plugin.sections,
)
if host_config.is_cluster
else get_section_kwargs(
parsed_sections_broker,
HostKey(host_config.hostname, ipaddress, source_type),
plugin.sections,
)
)
if not kwargs and not service.check_plugin_name.is_management_name():
# in 1.6 some plugins where discovered for management boards, but with
# the regular host plugins name. In this case retry with the source type
# forced to MANAGEMENT:
kwargs = (
get_section_cluster_kwargs(
parsed_sections_broker,
config_cache.get_clustered_service_node_keys(
host_config.hostname,
SourceType.MANAGEMENT,
service.description,
)
or [],
plugin.sections,
)
if host_config.is_cluster
else get_section_kwargs(
parsed_sections_broker,
HostKey(host_config.hostname, ipaddress, SourceType.MANAGEMENT),
plugin.sections,
)
)
if not kwargs: # no data found
return AggregatedResult(
submit=False,
data_received=False,
result=RECEIVED_NO_DATA,
cache_info=None,
)
kwargs = {
**kwargs,
**({} if service.item is None else {"item": service.item}),
**(
{}
if plugin.check_default_parameters is None
else {"params": _final_read_only_check_parameters(timespecific_parameters)}
),
}
with plugin_contexts.current_host(host_config.hostname), plugin_contexts.current_service(
service
), value_store_manager.namespace(service.id()):
result = _aggregate_results(check_function(**kwargs))
except (item_state.MKCounterWrapped, checking_classes.IgnoreResultsError) as e:
msg = str(e) or "No service summary available"
return AggregatedResult(
submit=False,
data_received=True,
result=(0, msg, []),
cache_info=None,
)
except MKTimeout:
raise
except Exception:
if cmk.utils.debug.enabled():
raise
table = check_table.get_check_table(host_config.hostname, skip_autochecks=True)
result = (
3,
cmk.base.crash_reporting.create_check_crash_dump(
host_name=host_config.hostname,
service_name=service.description,
plugin_name=service.check_plugin_name,
plugin_kwargs=globals().get("kwargs", {}),
is_manual=service.id() in table,
),
[],
)
return AggregatedResult(
submit=True,
data_received=True,
result=result,
cache_info=parsed_sections_broker.get_cache_info(plugin.sections),
)
def do_check(
hostname: HostName,
ipaddress: Optional[HostAddress],
*,
# The following arguments *must* remain optional for Nagios and the `DiscoCheckExecutor`.
# See Also: `cmk.base.discovery.check_discovery()`
fetcher_messages: Sequence[FetcherMessage] = (),
run_only_plugin_names: Optional[Set[CheckPluginName]] = None,
selected_sections: SectionNameCollection = NO_SELECTION,
dry_run: bool = False,
show_perfdata: bool = False,
) -> Tuple[int, List[ServiceDetails], List[ServiceAdditionalDetails],
List[str]]:
console.verbose("Checkmk version %s\n", cmk_version.__version__)
config_cache = config.get_config_cache()
host_config = config_cache.get_host_config(hostname)
exit_spec = host_config.exit_code_spec()
mode = Mode.CHECKING if selected_sections is NO_SELECTION else Mode.FORCE_SECTIONS
status: ServiceState = 0
infotexts: List[ServiceDetails] = []
long_infotexts: List[ServiceAdditionalDetails] = []
perfdata: List[str] = []
try:
license_usage.try_history_update()
# In case of keepalive we always have an ipaddress (can be 0.0.0.0 or :: when
# address is unknown). When called as non keepalive ipaddress may be None or
# is already an address (2nd argument)
if ipaddress is None and not host_config.is_cluster:
ipaddress = config.lookup_ip_address(host_config)
# When monitoring Checkmk clusters, the cluster nodes are responsible for fetching all
# information from the monitored host and cache the result for the cluster checks to be
# performed on the cached information.
#
# This means that in case of SNMP nodes, they need to take the clustered services of the
# node into account, fetch the needed sections and cache them for the cluster host.
#
# But later, when checking the node services, the node has to only deal with the unclustered
# services.
#
# TODO: clean this up. The fetched sections are computed in the checkers
# _make_configured_snmp_sections now.
#
belongs_to_cluster = len(config_cache.clusters_of(hostname)) > 0
services_to_fetch = _get_services_to_fetch(
host_name=hostname,
belongs_to_cluster=belongs_to_cluster,
config_cache=config_cache,
)
services_to_check = _filter_clustered_services(
config_cache=config_cache,
host_name=hostname,
belongs_to_cluster=belongs_to_cluster,
services=services_to_fetch,
run_only_plugin_names=run_only_plugin_names,
)
nodes = sources.make_nodes(
config_cache,
host_config,
ipaddress,
mode,
sources.make_sources(
host_config,
ipaddress,
mode=mode,
selected_sections=selected_sections,
),
)
if not fetcher_messages:
# Note: `fetch_all(sources)` is almost always called in similar
# code in discovery and inventory. The only other exception
# is `cmk.base.discovery.check_discovery(...)`. This does
# not seem right.
fetcher_messages = list(
sources.fetch_all(
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
))
with CPUTracker() as tracker:
broker = ParsedSectionsBroker()
result = sources.update_host_sections(
broker,
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
fetcher_messages=fetcher_messages,
selected_sections=selected_sections,
)
num_success, plugins_missing_data = _do_all_checks_on_host(
config_cache,
host_config,
ipaddress,
parsed_sections_broker=broker,
services=services_to_check,
dry_run=dry_run,
show_perfdata=show_perfdata,
)
if run_only_plugin_names is None:
inventory.do_inventory_actions_during_checking_for(
config_cache,
host_config,
ipaddress,
parsed_sections_broker=broker,
)
for source, host_sections in result:
source_state, source_output = source.summarize(host_sections)
if source_output != "":
status = max(status, source_state)
infotexts.append("[%s] %s" % (source.id, source_output))
if plugins_missing_data:
missing_data_status, missing_data_infotext = _check_plugins_missing_data(
plugins_missing_data,
exit_spec,
bool(num_success),
)
status = max(status, missing_data_status)
infotexts.append(missing_data_infotext)
total_times = tracker.duration
for msg in fetcher_messages:
total_times += msg.stats.duration
infotexts.append("execution time %.1f sec" %
total_times.process.elapsed)
if config.check_mk_perfdata_with_times:
perfdata += [
"execution_time=%.3f" % total_times.process.elapsed,
"user_time=%.3f" % total_times.process.user,
"system_time=%.3f" % total_times.process.system,
"children_user_time=%.3f" % total_times.process.children_user,
"children_system_time=%.3f" %
total_times.process.children_system,
]
summary: DefaultDict[str, Snapshot] = defaultdict(Snapshot.null)
for msg in fetcher_messages if fetcher_messages else ():
if msg.fetcher_type in (
FetcherType.PIGGYBACK,
FetcherType.PROGRAM,
FetcherType.SNMP,
FetcherType.TCP,
):
summary[{
FetcherType.PIGGYBACK: "agent",
FetcherType.PROGRAM: "ds",
FetcherType.SNMP: "snmp",
FetcherType.TCP: "agent",
}[msg.fetcher_type]] += msg.stats.duration
for phase, duration in summary.items():
perfdata.append("cmk_time_%s=%.3f" % (phase, duration.idle))
else:
perfdata.append("execution_time=%.3f" %
total_times.process.elapsed)
return status, infotexts, long_infotexts, perfdata
finally:
_submit_to_core.finalize()
def _cluster_scenario(monkeypatch):
hostname = "test-clusterhost"
ipaddress = "1.2.3.4"
node1_hostname = 'test-node1'
node2_hostname = 'test-node2'
def fake_lookup_ip_address(*_a, **_kw):
return ipaddress
monkeypatch.setattr(ip_lookup, "lookup_ip_address", fake_lookup_ip_address)
ts = Scenario()
ts.add_host(node1_hostname)
ts.add_host(node2_hostname)
ts.add_cluster(hostname, nodes=[node1_hostname, node2_hostname])
ts.set_ruleset("inventory_df_rules", [{
'value': {
'ignore_fs_types': ['tmpfs', 'nfs', 'smbfs', 'cifs', 'iso9660'],
'never_ignore_mountpoints': ['~.*/omd/sites/[^/]+/tmp$']
},
'condition': {
'host_labels': {
'cmk/check_mk_server': 'yes'
}
}
}])
ts.set_ruleset("clustered_services", [([], [node1_hostname], ['fs_'])])
host_config = ts.apply(monkeypatch).get_host_config(hostname)
DiscoveredHostLabelsStore(node1_hostname).save({
'node1_existing_label': {
'plugin_name': 'node1_plugin',
'value': 'true',
}
})
DiscoveredHostLabelsStore(hostname).save({
'existing_label': {
'plugin_name': 'foo',
'value': 'bar',
},
'another_label': {
'plugin_name': 'labels',
'value': 'true',
}
})
broker = ParsedSectionsBroker({
HostKey(hostname=node1_hostname,
ipaddress=ipaddress,
source_type=SourceType.HOST):
SectionsParser(host_sections=AgentHostSections(sections={
SectionName("labels"): [
[
'{"cmk/check_mk_server":"yes"}',
],
],
SectionName("df"): [
[
'/dev/sda1',
'vfat',
'523248',
'3668',
'519580',
'1%',
'/boot/test-efi',
],
[
'tmpfs',
'tmpfs',
'8152916',
'244',
'8152672',
'1%',
'/opt/omd/sites/test-heute/tmp',
],
],
}, ), ),
HostKey(hostname=node2_hostname,
ipaddress=ipaddress,
source_type=SourceType.HOST):
SectionsParser(host_sections=AgentHostSections(sections={
SectionName("labels"): [
[
'{"node2_live_label":"true"}',
],
],
SectionName("df"): [
[
'/dev/sda1',
'vfat',
'523248',
'3668',
'519580',
'1%',
'/boot/test-efi',
],
[
'tmpfs',
'tmpfs',
'8152916',
'244',
'8152672',
'1%',
'/opt/omd/sites/test-heute2/tmp',
],
],
}, ), ),
})
return ClusterScenario(
host_config,
ipaddress,
broker,
node1_hostname,
node2_hostname,
)
def _discover_host_labels_for_source_type(
*,
host_key: HostKey,
parsed_sections_broker: ParsedSectionsBroker,
discovery_parameters: DiscoveryParameters,
) -> Mapping[str, HostLabel]:
try:
host_data = parsed_sections_broker[host_key]
except KeyError:
return {}
host_labels = {}
try:
# We do *not* process all available raw sections. Instead we see which *parsed*
# sections would result from them, and then process those.
parse_sections = {
agent_based_register.get_section_plugin(rs).parsed_section_name
for rs in host_data.sections
}
applicable_sections = parsed_sections_broker.determine_applicable_sections(
parse_sections,
host_key.source_type,
)
console.vverbose("Trying host label discovery with: %s\n" %
", ".join(str(s.name) for s in applicable_sections))
for section_plugin in _sort_sections_by_label_priority(
applicable_sections):
kwargs = {
'section':
parsed_sections_broker.get_parsed_section(
host_key, section_plugin.parsed_section_name),
}
host_label_params = config.get_host_label_parameters(
host_key.hostname, section_plugin)
if host_label_params is not None:
kwargs["params"] = host_label_params
try:
for label in section_plugin.host_label_function(**kwargs):
console.vverbose(
f" {label.name}: {label.value} ({section_plugin.name})\n"
)
host_labels[label.name] = HostLabel(
label.name,
label.value,
section_plugin.name,
)
except (KeyboardInterrupt, MKTimeout):
raise
except Exception as exc:
if cmk.utils.debug.enabled(
) or discovery_parameters.on_error == "raise":
raise
if discovery_parameters.on_error == "warn":
console.error("Host label discovery of '%s' failed: %s\n" %
(section_plugin.name, exc))
except KeyboardInterrupt:
raise MKGeneralException("Interrupted by Ctrl-C.")
return host_labels
def get_check_preview(
*,
host_name: HostName,
max_cachefile_age: int,
use_cached_snmp_data: bool,
on_error: str,
) -> Tuple[CheckPreviewTable, DiscoveredHostLabels]:
"""Get the list of service of a host or cluster and guess the current state of
all services if possible"""
config_cache = config.get_config_cache()
host_config = config_cache.get_host_config(host_name)
ip_address = None if host_config.is_cluster else config.lookup_ip_address(host_config)
discovery_parameters = DiscoveryParameters(
on_error=on_error,
load_labels=False,
save_labels=False,
only_host_labels=False,
)
_set_cache_opts_of_checkers(use_cached_snmp_data=use_cached_snmp_data)
nodes = sources.make_nodes(
config_cache, host_config, ip_address, Mode.DISCOVERY,
sources.make_sources(
host_config,
ip_address,
mode=Mode.DISCOVERY,
on_scan_error=on_error,
))
parsed_sections_broker = ParsedSectionsBroker()
sources.update_host_sections(
parsed_sections_broker,
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
fetcher_messages=list(
sources.fetch_all(
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
)),
selected_sections=NO_SELECTION,
)
grouped_services, host_label_result = _get_host_services(
host_config,
ip_address,
parsed_sections_broker,
discovery_parameters,
)
table: CheckPreviewTable = []
for check_source, services_with_nodes in grouped_services.items():
for service, found_on_nodes in services_with_nodes:
plugin = agent_based_register.get_check_plugin(service.check_plugin_name)
params = _preview_params(host_name, service, plugin, check_source)
if check_source in ['legacy', 'active', 'custom']:
exitcode = None
output = u"WAITING - %s check, cannot be done offline" % check_source.title()
ruleset_name: Optional[RulesetName] = None
else:
ruleset_name = (str(plugin.check_ruleset_name)
if plugin and plugin.check_ruleset_name else None)
wrapped_params = (Parameters(wrap_parameters(params)) if plugin and
plugin.check_default_parameters is not None else None)
exitcode, output, _perfdata = checking.get_aggregated_result(
parsed_sections_broker,
host_config,
ip_address,
service,
plugin,
lambda p=wrapped_params: p, # type: ignore[misc] # "type of lambda"
).result
# Service discovery never uses the perfdata in the check table. That entry
# is constantly discarded, yet passed around(back and forth) as part of the
# discovery result in the request elements. Some perfdata VALUES are not parsable
# by ast.literal_eval such as "inf" it lead to ValueErrors. Thus keep perfdata empty
perfdata: List[MetricTuple] = []
table.append((
_preview_check_source(host_name, service, check_source),
str(service.check_plugin_name),
ruleset_name,
service.item,
service.parameters,
params,
service.description,
exitcode,
output,
perfdata,
service.service_labels.to_dict(),
found_on_nodes,
))
return table, DiscoveredHostLabels(
*{
# TODO (mo): According to unit tests, this is what was done prior to refactoring.
# Im not sure this is desired. If it is, it should be explained.
**{l.name: l for l in host_label_result.vanished},
**{l.name: l for l in host_label_result.present},
}.values())
def get_aggregated_result(
parsed_sections_broker: ParsedSectionsBroker,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
service: Service,
plugin: Optional[checking_classes.CheckPlugin],
timespecific_parameters: Union[LegacyCheckParameters,
TimespecificParameters],
*,
value_store_manager: value_store.ValueStoreManager,
persist_value_store_changes: bool,
) -> AggregatedResult:
"""Run the check function and aggregate the subresults
This function is also called during discovery.
"""
if plugin is None:
return AggregatedResult(
submit=True,
data_received=True,
result=ServiceCheckResult.check_not_implemented(),
cache_info=None,
)
config_cache = config.get_config_cache()
check_function = (_cluster_modes.get_cluster_check_function(
*config_cache.get_clustered_service_configuration(
host_config.hostname,
service.description,
),
plugin=plugin,
service_id=service.id(),
persist_value_store_changes=persist_value_store_changes,
) if host_config.is_cluster else plugin.check_function)
section_kws, error_result = _get_monitoring_data_kwargs_handle_pre20_services(
parsed_sections_broker,
host_config,
config_cache,
ipaddress,
service,
plugin.sections,
)
if not section_kws: # no data found
return AggregatedResult(
submit=False,
data_received=False,
result=error_result,
cache_info=None,
)
item_kw = {} if service.item is None else {"item": service.item}
params_kw = ({} if plugin.check_default_parameters is None else {
"params":
_final_read_only_check_parameters(timespecific_parameters)
})
try:
with plugin_contexts.current_host(
host_config.hostname), plugin_contexts.current_service(
service), value_store_manager.namespace(service.id()):
result = _aggregate_results(
check_function(
**item_kw,
**params_kw,
**section_kws,
))
except (item_state.MKCounterWrapped,
checking_classes.IgnoreResultsError) as e:
msg = str(e) or "No service summary available"
return AggregatedResult(
submit=False,
data_received=True,
result=ServiceCheckResult(output=msg),
cache_info=None,
)
except MKTimeout:
raise
except Exception:
if cmk.utils.debug.enabled():
raise
table = check_table.get_check_table(host_config.hostname,
skip_autochecks=True)
result = ServiceCheckResult(
3,
cmk.base.crash_reporting.create_check_crash_dump(
host_name=host_config.hostname,
service_name=service.description,
plugin_name=service.check_plugin_name,
plugin_kwargs={
**item_kw,
**params_kw,
**section_kws
},
is_manual=service.id() in table,
),
)
return AggregatedResult(
submit=True,
data_received=True,
result=result,
cache_info=parsed_sections_broker.get_cache_info(plugin.sections),
)
def do_discovery(
arg_hostnames: Set[HostName],
*,
selected_sections: SectionNameCollection,
run_only_plugin_names: Optional[Set[CheckPluginName]],
arg_only_new: bool,
only_host_labels: bool = False,
) -> None:
config_cache = config.get_config_cache()
use_caches = not arg_hostnames or cmk.core_helpers.cache.FileCacheFactory.maybe
on_error = "raise" if cmk.utils.debug.enabled() else "warn"
discovery_parameters = DiscoveryParameters(
on_error=on_error,
load_labels=arg_only_new,
save_labels=True,
only_host_labels=only_host_labels,
)
host_names = _preprocess_hostnames(arg_hostnames, config_cache, only_host_labels)
mode = Mode.DISCOVERY if selected_sections is NO_SELECTION else Mode.FORCE_SECTIONS
# Now loop through all hosts
for host_name in sorted(host_names):
host_config = config_cache.get_host_config(host_name)
section.section_begin(host_name)
try:
ipaddress = config.lookup_ip_address(host_config)
nodes = sources.make_nodes(
config_cache,
host_config,
ipaddress,
mode,
sources.make_sources(
host_config,
ipaddress,
mode=mode,
selected_sections=selected_sections,
on_scan_error=on_error,
),
)
max_cachefile_age = config.discovery_max_cachefile_age() if use_caches else 0
parsed_sections_broker = ParsedSectionsBroker()
sources.update_host_sections(
parsed_sections_broker,
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
fetcher_messages=list(
sources.fetch_all(
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
)),
selected_sections=selected_sections,
)
_do_discovery_for(
host_name,
ipaddress,
parsed_sections_broker,
run_only_plugin_names,
arg_only_new,
discovery_parameters,
)
except Exception as e:
if cmk.utils.debug.enabled():
raise
section.section_error("%s" % e)
finally:
cmk.utils.cleanup.cleanup_globals()
def _do_inv_for_realhost(
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
*,
parsed_sections_broker: ParsedSectionsBroker,
run_plugin_names: Container[InventoryPluginName],
retentions_tracker: RetentionsTracker,
) -> InventoryTrees:
tree_aggregator = TreeAggregator()
_set_cluster_property(tree_aggregator.trees.inventory, host_config)
section.section_step("Executing inventory plugins")
for inventory_plugin in agent_based_register.iter_all_inventory_plugins():
if inventory_plugin.name not in run_plugin_names:
continue
for source_type in (SourceType.HOST, SourceType.MANAGEMENT):
kwargs = get_section_kwargs(
parsed_sections_broker,
HostKey(host_config.hostname, ipaddress, source_type),
inventory_plugin.sections,
)
if not kwargs:
console.vverbose(
" %s%s%s%s: skipped (no data)\n",
tty.yellow,
tty.bold,
inventory_plugin.name,
tty.normal,
)
continue
# Inventory functions can optionally have a second argument: parameters.
# These are configured via rule sets (much like check parameters).
if inventory_plugin.inventory_ruleset_name is not None:
kwargs = {
**kwargs,
"params":
host_config.inventory_parameters(
inventory_plugin.inventory_ruleset_name),
}
exception = tree_aggregator.aggregate_results(
inventory_generator=inventory_plugin.inventory_function(
**kwargs),
retentions_tracker=retentions_tracker,
raw_cache_info=parsed_sections_broker.get_cache_info(
inventory_plugin.sections),
is_legacy_plugin=inventory_plugin.module is None,
)
if exception:
console.warning(
" %s%s%s%s: failed: %s",
tty.red,
tty.bold,
inventory_plugin.name,
tty.normal,
exception,
)
else:
console.verbose(" %s%s%s%s", tty.green, tty.bold,
inventory_plugin.name, tty.normal)
console.vverbose(": ok\n")
console.verbose("\n")
return tree_aggregator.trees
def discover_on_host(
*,
config_cache: config.ConfigCache,
host_config: config.HostConfig,
mode: str,
service_filters: Optional[_ServiceFilters],
on_error: str,
use_cached_snmp_data: bool,
max_cachefile_age: int,
) -> DiscoveryResult:
console.verbose(" Doing discovery with mode '%s'...\n" % mode)
host_name = host_config.hostname
result = DiscoveryResult()
discovery_parameters = DiscoveryParameters(
on_error=on_error,
load_labels=(mode != "remove"),
save_labels=(mode != "remove"),
only_host_labels=(mode == "only-host-labels"),
)
if host_name not in config_cache.all_active_hosts():
result.error_text = ""
return result
_set_cache_opts_of_checkers(use_cached_snmp_data=use_cached_snmp_data)
try:
# in "refresh" mode we first need to remove all previously discovered
# checks of the host, so that _get_host_services() does show us the
# new discovered check parameters.
if mode == "refresh":
result.self_removed += host_config.remove_autochecks() # this is cluster-aware!
if host_config.is_cluster:
ipaddress = None
else:
ipaddress = config.lookup_ip_address(host_config)
nodes = sources.make_nodes(
config_cache,
host_config,
ipaddress,
Mode.DISCOVERY,
sources.make_sources(
host_config,
ipaddress,
mode=Mode.DISCOVERY,
on_scan_error=on_error,
),
)
parsed_sections_broker = ParsedSectionsBroker()
sources.update_host_sections(
parsed_sections_broker,
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
fetcher_messages=list(
sources.fetch_all(
nodes,
max_cachefile_age=max_cachefile_age,
host_config=host_config,
)),
selected_sections=NO_SELECTION,
)
# Compute current state of new and existing checks
services, host_labels = _get_host_services(
host_config,
ipaddress,
parsed_sections_broker,
discovery_parameters,
)
old_services = services.get("old", [])
# Create new list of checks
new_services = _get_post_discovery_services(host_name, services, service_filters or
_ServiceFilters.accept_all(), result, mode)
host_config.set_autochecks(new_services)
result.diff_text = make_object_diff(
_make_services_audit_log_object([x.service for x in old_services]),
_make_services_audit_log_object([x.service for x in new_services]))
except MKTimeout:
raise # let general timeout through
except Exception as e:
if cmk.utils.debug.enabled():
raise
result.error_text = str(e)
else:
if mode != "remove":
result.self_new_host_labels = len(host_labels.new)
result.self_total_host_labels = len(host_labels.present)
result.self_total = result.self_new + result.self_kept
return result
