def _all_sources_fail(
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
) -> bool:
"""We want to check if ALL data sources of a host fail:
By default a host has the auto-piggyback data source. We remove it if
it's not a pure piggyback host and there's no piggyback data available
for this host.
In this case the piggyback data source never fails (self._exception = None)."""
if host_config.is_cluster:
return False
# TODO(ml): This function makes no sense and is no op anyway.
# We could fix it by actually searching for errors in the sources
# as it seems that it is what was meant initially.
exceptions_by_source = {
source.id: None
for source in checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.INVENTORY,
)
}
if "piggyback" in exceptions_by_source and not len(exceptions_by_source) == 1\
and not host_config.has_piggyback_data:
del exceptions_by_source["piggyback"]
return all(exception is not None
for exception in exceptions_by_source.values())
def _get_multi_host_sections_for_inv(
config_cache: config.ConfigCache,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
) -> Tuple[MultiHostSections, Sequence[Tuple[ABCSource, result.Result[
ABCHostSections, Exception]]]]:
if host_config.is_cluster:
return MultiHostSections(), []
sources = checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.INVENTORY,
)
multi_host_sections = MultiHostSections()
results = checkers.update_host_sections(
multi_host_sections,
checkers.make_nodes(
config_cache,
host_config,
ipaddress,
checkers.Mode.INVENTORY,
sources,
),
max_cachefile_age=host_config.max_cachefile_age,
selected_raw_sections=None,
host_config=host_config,
)
return multi_host_sections, results
def test_host_config_creates_passing_source_sources(
monkeypatch,
hostname,
mode,
tags,
sources,
):
ts = make_scenario(hostname, tags)
ts.apply(monkeypatch)
host_config = config.HostConfig.make_host_config(hostname)
ipaddress = "127.0.0.1"
assert [type(c) for c in make_sources(
host_config,
ipaddress,
mode=mode,
)] == sources
def _fetch_multi_host_sections_for_inv(
config_cache: config.ConfigCache,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
selected_sections: checkers.SectionNameCollection,
) -> Tuple[MultiHostSections, Sequence[Tuple[Source, result.Result[
HostSections, Exception]]]]:
if host_config.is_cluster:
return MultiHostSections(), []
mode = (checkers.Mode.INVENTORY
if selected_sections is checkers.NO_SELECTION else
checkers.Mode.FORCE_SECTIONS)
sources = checkers.make_sources(
host_config,
ipaddress,
mode=mode,
selected_sections=selected_sections,
)
nodes = checkers.make_nodes(
config_cache,
host_config,
ipaddress,
mode,
sources,
)
multi_host_sections = MultiHostSections()
results = checkers.update_host_sections(
multi_host_sections,
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
fetcher_messages=list(
checkers.fetch_all(
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
)),
selected_sections=selected_sections,
)
return multi_host_sections, results
def _fetch_multi_host_sections_for_inv(
config_cache: config.ConfigCache,
host_config: config.HostConfig,
ipaddress: Optional[HostAddress],
) -> Tuple[MultiHostSections, Sequence[Tuple[Source, result.Result[
HostSections, Exception]]]]:
if host_config.is_cluster:
return MultiHostSections(), []
sources = checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.INVENTORY,
)
for source in sources:
_configure_source_for_inv(source)
nodes = checkers.make_nodes(
config_cache,
host_config,
ipaddress,
checkers.Mode.INVENTORY,
sources,
)
multi_host_sections = MultiHostSections()
results = checkers.update_host_sections(
multi_host_sections,
nodes,
max_cachefile_age=host_config.max_cachefile_age,
selected_raw_sections=None,
host_config=host_config,
fetcher_messages=list(
checkers.fetch_all(
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
selected_raw_sections=None,
)),
)
return multi_host_sections, results
def do_inv(hostnames: List[HostName]) -> None:
store.makedirs(cmk.utils.paths.inventory_output_dir)
store.makedirs(cmk.utils.paths.inventory_archive_dir)
config_cache = config.get_config_cache()
for hostname in hostnames:
section.section_begin(hostname)
try:
host_config = config.HostConfig.make_host_config(hostname)
if host_config.is_cluster:
ipaddress = None
else:
ipaddress = ip_lookup.lookup_ip_address(host_config)
inventory_tree, status_data_tree = _do_inv_for(
config_cache,
host_config,
ipaddress,
sources=checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.INVENTORY,
),
multi_host_sections=None,
)[:2]
_run_inventory_export_hooks(host_config, inventory_tree)
_show_inventory_results_on_console(inventory_tree,
status_data_tree)
except Exception as e:
if cmk.utils.debug.enabled():
raise
section.section_error("%s" % e)
finally:
cmk.utils.cleanup.cleanup_globals()
def do_check(
hostname: HostName,
ipaddress: Optional[HostAddress],
only_check_plugin_names: Optional[Set[CheckPluginName]] = None,
fetcher_messages: Optional[Sequence[FetcherMessage]] = None
) -> 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()
status: ServiceState = 0
infotexts: List[ServiceDetails] = []
long_infotexts: List[ServiceAdditionalDetails] = []
perfdata: List[str] = []
try:
with cpu_tracking.execute(), cpu_tracking.phase("busy"):
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 = ip_lookup.lookup_ip_address(host_config)
item_state.load(hostname)
# 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.
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,
only_check_plugins=only_check_plugin_names,
)
services_to_check = _filter_clustered_services(
config_cache=config_cache,
host_name=hostname,
belongs_to_cluster=belongs_to_cluster,
services=services_to_fetch,
)
# see which raw sections we may need
selected_raw_sections = agent_based_register.get_relevant_raw_sections(
check_plugin_names=(s.check_plugin_name for s in services_to_fetch),
consider_inventory_plugins=host_config.do_status_data_inventory,
)
sources = checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.CHECKING,
)
mhs = MultiHostSections()
result = checkers.update_host_sections(
mhs,
checkers.make_nodes(
config_cache,
host_config,
ipaddress,
checkers.Mode.CHECKING,
sources,
),
selected_raw_sections=selected_raw_sections,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
fetcher_messages=fetcher_messages,
)
num_success, plugins_missing_data = _do_all_checks_on_host(
config_cache,
host_config,
ipaddress,
multi_host_sections=mhs,
services=services_to_check,
only_check_plugins=only_check_plugin_names,
)
inventory.do_inventory_actions_during_checking_for(
config_cache,
host_config,
ipaddress,
sources=sources,
multi_host_sections=mhs,
)
if _submit_to_core:
item_state.save(hostname)
for source, host_sections in result:
source_state, source_output, source_perfdata = source.summarize(host_sections)
if source_output != "":
status = max(status, source_state)
infotexts.append("[%s] %s" % (source.id, source_output))
perfdata.extend([_convert_perf_data(p) for p in source_perfdata])
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)
for msg in fetcher_messages if fetcher_messages else ():
cpu_tracking.update(msg.stats.cpu_times)
phase_times = cpu_tracking.get_times()
total_times = phase_times["busy"]
infotexts.append("execution time %.1f sec" % total_times.run_time)
if config.check_mk_perfdata_with_times:
perfdata += [
"execution_time=%.3f" % total_times.run_time,
"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,
]
for phase, times in phase_times.items():
if phase in ["agent", "snmp", "ds"]:
t = times.run_time - sum(times.process[:4]) # real time - CPU time
perfdata.append("cmk_time_%s=%.3f" % (phase, t))
else:
perfdata.append("execution_time=%.3f" % total_times.run_time)
return status, infotexts, long_infotexts, perfdata
finally:
if _checkresult_file_fd is not None:
_close_checkresult_file()
# "ipaddress is not None": At least when working with a cluster host it seems the ipaddress
# may be None. This needs to be understood in detail and cleaned up. As the InlineSNMP
# stats feature is a very rarely used debugging feature, the analyzation and fix is
# postponed now.
if config.record_inline_snmp_stats and ipaddress is not None and host_config.snmp_config(
ipaddress).snmp_backend == "inline":
inline.snmp_stats_save()
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: checkers.SectionNameCollection = checkers.NO_SELECTION,
submit_to_core: bool = True,
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 = checkers.Mode.CHECKING if selected_sections is checkers.NO_SELECTION else checkers.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 = ip_lookup.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 = checkers.make_nodes(
config_cache,
host_config,
ipaddress,
mode,
checkers.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(
checkers.fetch_all(
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
))
with CPUTracker() as tracker:
mhs = MultiHostSections()
result = checkers.update_host_sections(
mhs,
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,
multi_host_sections=mhs,
services=services_to_check,
submit_to_core=submit_to_core,
show_perfdata=show_perfdata,
)
if run_only_plugin_names is None:
inventory.do_inventory_actions_during_checking_for(
config_cache,
host_config,
ipaddress,
multi_host_sections=mhs,
)
for source, host_sections in result:
source_state, source_output, source_perfdata = source.summarize(
host_sections)
if source_output != "":
status = max(status, source_state)
infotexts.append("[%s] %s" % (source.id, source_output))
perfdata.extend(
[_convert_perf_data(p) for p in source_perfdata])
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:
if _checkresult_file_fd is not None:
_close_checkresult_file()
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 lookup_ip_address(host_config, family=None, for_mgmt_board=False):
return hosts[host_config.hostname]
def check(_, *args, **kwargs):
return result.OK(AgentHostSections(sections={section_name: [[str(section_name)]]}))
monkeypatch.setattr(
ip_lookup,
"lookup_ip_address",
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)
)
mhs = MultiHostSections()
update_host_sections(
mhs,
nodes,
max_cachefile_age=host_config.max_cachefile_age,
host_config=host_config,
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,
)
assert len(mhs) == 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 mhs
section = mhs[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 do_inv_check(
hostname: HostName, options: Dict[str, int]
) -> Tuple[ServiceState, List[ServiceDetails], List[ServiceAdditionalDetails],
List[MetricTuple]]:
_inv_hw_changes = options.get("hw-changes", 0)
_inv_sw_changes = options.get("sw-changes", 0)
_inv_sw_missing = options.get("sw-missing", 0)
_inv_fail_status = options.get("inv-fail-status", 1)
config_cache = config.get_config_cache()
host_config = config.HostConfig.make_host_config(hostname)
if host_config.is_cluster:
ipaddress = None
else:
ipaddress = ip_lookup.lookup_ip_address(host_config)
status = 0
infotexts: List[str] = []
long_infotexts: List[str] = []
sources = checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.INVENTORY,
)
inventory_tree, status_data_tree, results = _do_inv_for(
config_cache,
host_config,
ipaddress,
sources=sources,
multi_host_sections=None,
)
#TODO add cluster if and only if all sources do not fail?
if _all_sources_fail(host_config, ipaddress):
old_tree, sources_state = None, 1
status = max(status, sources_state)
infotexts.append("Cannot update tree%s" %
check_api_utils.state_markers[sources_state])
else:
old_tree = _save_inventory_tree(hostname, inventory_tree)
_run_inventory_export_hooks(host_config, inventory_tree)
if inventory_tree.is_empty() and status_data_tree.is_empty():
infotexts.append("Found no data")
else:
infotexts.append("Found %d inventory entries" %
inventory_tree.count_entries())
# Node 'software' is always there because _do_inv_for creates this node for cluster info
if not inventory_tree.get_sub_container(['software']).has_edge('packages')\
and _inv_sw_missing:
infotexts.append("software packages information is missing" +
check_api_utils.state_markers[_inv_sw_missing])
status = max(status, _inv_sw_missing)
if old_tree is not None:
if not old_tree.is_equal(inventory_tree, edges=["software"]):
infotext = "software changes"
if _inv_sw_changes:
status = max(status, _inv_sw_changes)
infotext += check_api_utils.state_markers[_inv_sw_changes]
infotexts.append(infotext)
if not old_tree.is_equal(inventory_tree, edges=["hardware"]):
infotext = "hardware changes"
if _inv_hw_changes:
status = max(status, _inv_hw_changes)
infotext += check_api_utils.state_markers[_inv_hw_changes]
infotexts.append(infotext)
if not status_data_tree.is_empty():
infotexts.append("Found %s status entries" %
status_data_tree.count_entries())
for source, host_sections in results:
source_state, source_output, _source_perfdata = source.summarize(
host_sections)
if source_state != 0:
# Do not output informational things (state == 0). Also do not use source states
# which would overwrite "State when inventory fails" in the ruleset
# "Do hardware/software Inventory".
# These information and source states are handled by the "Check_MK" service
status = max(_inv_fail_status, status)
infotexts.append("[%s] %s" % (source.id, source_output))
return status, infotexts, long_infotexts, []
def dump_host(hostname: HostName) -> None:
config_cache = config.get_config_cache()
host_config = config_cache.get_host_config(hostname)
out.output("\n")
if host_config.is_cluster:
nodes = host_config.nodes
if nodes is None:
raise RuntimeError()
color = tty.bgmagenta
add_txt = " (cluster of " + (", ".join(nodes)) + ")"
else:
color = tty.bgblue
add_txt = ""
out.output("%s%s%s%-78s %s\n" %
(color, tty.bold, tty.white, hostname + add_txt, tty.normal))
ipaddress = _ip_address_for_dump_host(host_config)
addresses: Optional[str] = ""
if not host_config.is_ipv4v6_host:
addresses = ipaddress
else:
try:
if host_config.is_ipv6_primary:
secondary = _ip_address_for_dump_host(host_config, 4)
else:
secondary = _ip_address_for_dump_host(host_config, 6)
except Exception:
secondary = "X.X.X.X"
addresses = "%s, %s" % (ipaddress, secondary)
if host_config.is_ipv6_primary:
addresses += " (Primary: IPv6)"
else:
addresses += " (Primary: IPv4)"
out.output(tty.yellow + "Addresses: " + tty.normal +
(addresses if addresses is not None else "No IP") + "\n")
tag_template = tty.bold + "[" + tty.normal + "%s" + tty.bold + "]" + tty.normal
tags = [(tag_template % ":".join(t))
for t in sorted(host_config.tag_groups.items())]
out.output(tty.yellow + "Tags: " + tty.normal +
", ".join(tags) + "\n")
labels = [
tag_template % ":".join(l) for l in sorted(host_config.labels.items())
]
out.output(tty.yellow + "Labels: " + tty.normal +
", ".join(labels) + "\n")
# TODO: Clean this up once cluster parent handling has been moved to HostConfig
if host_config.is_cluster:
parents_list = host_config.nodes
if parents_list is None:
raise RuntimeError()
else:
parents_list = host_config.parents
if len(parents_list) > 0:
out.output(tty.yellow + "Parents: " + tty.normal +
", ".join(parents_list) + "\n")
out.output(tty.yellow + "Host groups: " + tty.normal +
", ".join(host_config.hostgroups) + "\n")
out.output(tty.yellow + "Contact groups: " + tty.normal +
", ".join(host_config.contactgroups) + "\n")
agenttypes = [
source.description for source in checkers.make_sources(
host_config,
ipaddress,
mode=checkers.Mode.NONE,
)
]
if host_config.is_ping_host:
agenttypes.append('PING only')
out.output(tty.yellow + "Agent mode: " + tty.normal)
out.output(host_config.agent_description + "\n")
out.output(tty.yellow + "Type of agent: " + tty.normal)
if len(agenttypes) == 1:
out.output(agenttypes[0] + "\n")
else:
out.output("\n ")
out.output("\n ".join(agenttypes) + "\n")
out.output(tty.yellow + "Services:" + tty.normal + "\n")
headers = ["checktype", "item", "params", "description", "groups"]
colors = [tty.normal, tty.blue, tty.normal, tty.green, tty.normal]
table_data = []
for service in sorted(check_table.get_check_table(hostname).values(),
key=lambda s: s.description):
table_data.append([
str(service.check_plugin_name),
str(service.item),
_evaluate_params(service.parameters), service.description,
",".join(
config_cache.servicegroups_of_service(hostname,
service.description))
])
tty.print_table(headers, colors, table_data, " ")
