def host_context(host_name: HostName, *, write_state: bool):
"""Make a bit of context information globally available
So that functions called by checks know this context.
This is used for both legacy and agent_based API.
"""
# TODO: this is a mixture of legacy and new Check-API mechanisms. Clean this up!
try:
check_api_utils.set_hostname(host_name)
item_state.load(host_name)
yield
finally:
check_api_utils.reset_hostname()
if write_state:
item_state.save(host_name)
item_state.cleanup_item_states()
def current_host(host_name_: str, *, write_state: bool):
"""Make a bit of context information globally available
So that functions called by checks know this context.
This is used for both legacy and agent_based API.
"""
global _hostname
# we do not encourage nested contexts at all -- but it should work.
previous_host_name = _hostname
_hostname = host_name_
# TODO: this is a mixture of legacy and new Check-API mechanisms. Clean this up!
try:
item_state.load(host_name_)
yield
finally:
_hostname = previous_host_name
if write_state:
item_state.save()
item_state.cleanup_item_states()
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, ipaddress, only_check_plugin_names=None):
# type: (HostName, Optional[HostAddress], Optional[List[CheckPluginName]]) -> Tuple[int, List[ServiceDetails], List[ServiceAdditionalDetails], List[Text]]
cpu_tracking.start("busy")
console.verbose("Check_MK version %s\n", six.ensure_str(cmk.__version__))
config_cache = config.get_config_cache()
host_config = config_cache.get_host_config(hostname)
exit_spec = host_config.exit_code_spec()
status = 0 # type: ServiceState
infotexts = [] # type: List[ServiceDetails]
long_infotexts = [] # type: List[ServiceAdditionalDetails]
perfdata = [] # type: List[Text]
try:
# 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(hostname)
item_state.load(hostname)
sources = data_sources.DataSources(hostname, ipaddress)
num_success, missing_sections = \
_do_all_checks_on_host(sources, host_config, ipaddress, only_check_plugin_names)
if _submit_to_core:
item_state.save(hostname)
for source in sources.get_data_sources():
source_state, source_output, source_perfdata = source.get_summary_result_for_checking(
)
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 missing_sections and num_success > 0:
missing_sections_status, missing_sections_infotext = \
_check_missing_sections(missing_sections, exit_spec)
status = max(status, missing_sections_status)
infotexts.append(missing_sections_infotext)
elif missing_sections:
infotexts.append("Got no information from host")
status = max(status, cast(int, exit_spec.get("empty_output", 2)))
cpu_tracking.end()
phase_times = cpu_tracking.get_times()
total_times = phase_times["TOTAL"]
run_time = total_times[4]
infotexts.append("execution time %.1f sec" % run_time)
if config.check_mk_perfdata_with_times:
perfdata += [
"execution_time=%.3f" % run_time,
"user_time=%.3f" % total_times[0],
"system_time=%.3f" % total_times[1],
"children_user_time=%.3f" % total_times[2],
"children_system_time=%.3f" % total_times[3],
]
for phase, times in phase_times.items():
if phase in ["agent", "snmp", "ds"]:
t = times[4] - sum(times[:4]) # real time - CPU time
perfdata.append("cmk_time_%s=%.3f" % (phase, t))
else:
perfdata.append("execution_time=%.3f" % 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).is_inline_snmp_host:
inline_snmp.save_snmp_stats()
def do_check(
hostname: HostName,
ipaddress: Optional[HostAddress],
only_check_plugin_names: Optional[Set[CheckPluginName]] = None
) -> Tuple[int, List[ServiceDetails], List[ServiceAdditionalDetails], List[str]]:
cpu_tracking.start("busy")
console.verbose("Check_MK 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:
# 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(hostname)
item_state.load(hostname)
services = _get_filtered_services(
host_name=hostname,
belongs_to_cluster=len(config_cache.clusters_of(hostname)) > 0,
config_cache=config_cache,
only_check_plugins=only_check_plugin_names,
)
# see which raw sections we may need
selected_raw_sections = config.get_relevant_raw_sections(
CheckPluginName(maincheckify(s.check_plugin_name)) for s in services)
sources = data_sources.DataSources(
hostname,
ipaddress,
sources=data_sources.make_sources(
host_config,
ipaddress,
selected_raw_sections=selected_raw_sections,
),
)
num_success, plugins_missing_data = _do_all_checks_on_host(
services,
sources,
host_config,
ipaddress,
only_check_plugin_names,
)
if _submit_to_core:
item_state.save(hostname)
for source in sources:
source_state, source_output, source_perfdata = source.get_summary_result_for_checking()
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)
cpu_tracking.end()
phase_times = cpu_tracking.get_times()
total_times = phase_times["TOTAL"]
run_time = total_times[4]
infotexts.append("execution time %.1f sec" % run_time)
if config.check_mk_perfdata_with_times:
perfdata += [
"execution_time=%.3f" % run_time,
"user_time=%.3f" % total_times[0],
"system_time=%.3f" % total_times[1],
"children_user_time=%.3f" % total_times[2],
"children_system_time=%.3f" % total_times[3],
]
for phase, times in phase_times.items():
if phase in ["agent", "snmp", "ds"]:
t = times[4] - sum(times[:4]) # real time - CPU time
perfdata.append("cmk_time_%s=%.3f" % (phase, t))
else:
perfdata.append("execution_time=%.3f" % 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).is_inline_snmp_host:
inline.snmp_stats_save()
def do_check(
hostname: HostName,
ipaddress: Optional[HostAddress],
*,
# The next two *must* remain optional for Nagios and the `DiscoCheckExecutor`.
# See Also: `cmk.base.discovery.check_discovery()`
fetcher_messages: Sequence[FetcherMessage] = (),
only_check_plugin_names: Optional[Set[CheckPluginName]] = 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 CPUTracker() as tracker:
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,
)
nodes = checkers.make_nodes(
config_cache,
host_config,
ipaddress,
checkers.Mode.CHECKING,
sources,
)
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,
selected_raw_sections=selected_raw_sections,
))
mhs = MultiHostSections()
result = checkers.update_host_sections(
mhs,
nodes,
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)
total_times = tracker.duration
for msg in fetcher_messages if fetcher_messages else ():
total_times += msg.stats.cpu_tracker.duration
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,
]
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.cpu_tracker.duration
for phase, duration in summary.items():
t = duration.run_time - sum(
duration.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()
