Module: check_mk Branch: master Commit: a76dacd3167f3388c5f177e6949ee9883b14b0e2 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=a76dacd3167f33… Author: Andreas Boesl &lt;ab(a)mathias-kettner.de&gt; Date: Tue Mar 26 14:50:45 2019 +0100 WK7250: Check_MK Discovery: Clusters were unable to automatically discover new services Change-Id: Ie975f6998666c67ee598bface62457720fa34476 --- .werks/7250 | 22 ++++++++++++++ cmk_base/automations/check_mk.py | 4 ++- cmk_base/discovery.py | 64 ++++++++++++++++++++++++---------------- 3 files changed, 64 insertions(+), 26 deletions(-) diff --git a/.werks/7250 b/.werks/7250 new file mode 100644 index 0000000..fb19dbf --- /dev/null +++ b/.werks/7250 @@ -0,0 +1,22 @@ +Title: Check_MK Discovery: Clusters were unable to automatically discover new services +Level: 1 +Component: checks +Class: fix +Compatible: incomp +Edition: cre +State: unknown +Version: 1.6.0i1 +Date: 1553243799 + +The <tt>Check_MK Discovery</tt> check for a cluster was able to identify new/vanished services. +However, the automatic discovery mechanism in the background failed to integrate them into the configuration. + + +This has been fixed. There is still some constraint which makes this werk incompatible. +The <tt>Periodic service discovery</tt> rule allows you to configure <tt>Automatcally activate changes</tt> for the specified host. +If you want automatic activation of new services for cluster hosts, you have to configured a rule on the nodes instead of the cluster hosts. +So any <tt>Periodic service discovery</tt> rules of clusters should be duplicated onto the nodes. + + +This requirement is currently necessary because of the internal configuration. +Clusters never have an explicit configuration file, the actual clusters configuration is compute on demand. diff --git a/cmk_base/automations/check_mk.py b/cmk_base/automations/check_mk.py index b883ba6..9c4d799 100644 --- a/cmk_base/automations/check_mk.py +++ b/cmk_base/automations/check_mk.py @@ -116,7 +116,9 @@ class AutomationDiscovery(DiscoveryAutomation): for hostname in hostnames: result, error = discovery.discover_on_host(how, hostname, do_snmp_scan, use_caches, on_error) - counts[hostname] = result + counts[hostname] = result["self_new"], result["self_removed"], result[ + "self_kept"], result["self_total"] + if error is not None: failed_hosts[hostname] = error else: diff --git a/cmk_base/discovery.py b/cmk_base/discovery.py index f3a4abd..ae38c19 100644 --- a/cmk_base/discovery.py +++ b/cmk_base/discovery.py @@ -213,10 +213,17 @@ def discover_on_host(mode, use_caches, on_error="ignore", service_filter=None): - counts = {"added": 0, "removed": 0, "kept": 0} + counts = { + "self_new": 0, + "self_removed": 0, + "self_kept": 0, + "self_total": 0, + "clustered_new": 0, + "clustered_vanished": 0, + } if hostname not in config.all_active_realhosts(): - return [0, 0, 0, 0], "" + return counts, "" if service_filter is None: service_filter = lambda hostname, check_plugin_name, item: True @@ -228,7 +235,7 @@ def discover_on_host(mode, # checks of the host, so that _get_host_services() does show us the # new discovered check parameters. if mode == "refresh": - counts["removed"] += remove_autochecks_of(hostname) # this is cluster-aware! + counts["self_removed"] += remove_autochecks_of(hostname) # this is cluster-aware! if config.is_cluster(hostname): ipaddress = None @@ -259,13 +266,13 @@ def discover_on_host(mode, if check_source == "new": if mode in ("new", "fixall", "refresh") and service_filter( hostname, check_plugin_name, item): - counts["added"] += 1 + counts["self_new"] += 1 new_items[(check_plugin_name, item)] = paramstring elif check_source in ("old", "ignored"): # keep currently existing valid services in any case new_items[(check_plugin_name, item)] = paramstring - counts["kept"] += 1 + counts["self_kept"] += 1 elif check_source == "vanished": # keep item, if we are currently only looking for new services @@ -273,12 +280,13 @@ def discover_on_host(mode, if mode not in ("fixall", "remove") or not service_filter(hostname, check_plugin_name, item): new_items[(check_plugin_name, item)] = paramstring - counts["kept"] += 1 + counts["self_kept"] += 1 else: - counts["removed"] += 1 + counts["self_removed"] += 1 elif check_source.startswith("clustered_"): # Silently keep clustered services + counts[check_source] += 1 new_items[(check_plugin_name, item)] = paramstring else: @@ -292,8 +300,9 @@ def discover_on_host(mode, if cmk.utils.debug.enabled(): raise err = str(e) - return [counts["added"], counts["removed"], counts["kept"], - counts["added"] + counts["kept"]], err + + counts["self_total"] = counts["self_new"] + counts["self_kept"] + return counts, err #. @@ -397,8 +406,12 @@ def check_discovery(hostname, ipaddress): "ignored: %s: %s" % (check_plugin_name, config.service_description(hostname, check_plugin_name, item))) - _set_rediscovery_flag(hostname, need_rediscovery) if need_rediscovery: + if config.is_cluster(hostname): + for nodename in config.nodes_of(hostname): + _set_rediscovery_flag(nodename) + else: + _set_rediscovery_flag(hostname) infotexts.append("rediscovery scheduled") # Add data source information to check results @@ -436,7 +449,7 @@ def default_discovery_check_parameters(): } -def _set_rediscovery_flag(hostname, need_rediscovery): +def _set_rediscovery_flag(hostname): def touch(filename): if not os.path.exists(filename): f = open(filename, "w") @@ -444,16 +457,10 @@ def _set_rediscovery_flag(hostname, need_rediscovery): autodiscovery_dir = cmk.utils.paths.var_dir + '/autodiscovery' discovery_filename = os.path.join(autodiscovery_dir, hostname) - if need_rediscovery: - if not os.path.exists(autodiscovery_dir): - os.makedirs(autodiscovery_dir) - touch(discovery_filename) - else: - if os.path.exists(discovery_filename): - try: - os.remove(discovery_filename) - except OSError: - pass + + if not os.path.exists(autodiscovery_dir): + os.makedirs(autodiscovery_dir) + touch(discovery_filename) class DiscoveryTimeout(Exception): @@ -595,12 +602,19 @@ def _discover_marked_host(hostname, all_hosts, now_ts, oldest_queued): # compatible with the automation code console.verbose(" failed: host is offline\n") else: - new_services, removed_services, kept_services, total_services = result - if new_services == 0 and removed_services == 0 and kept_services == total_services: + if result["self_new"] == 0 and\ + result["self_removed"] == 0 and\ + result["self_kept"] == result["self_total"] and\ + result["clustered_new"] == 0 and\ + result["clustered_vanished"] == 0: console.verbose(" nothing changed.\n") else: - console.verbose( - " %d new, %d removed, %d kept, %d total services.\n" % (tuple(result))) + console.verbose(" %(self_new)s new, %(self_removed)s removed, "\ + "%(self_kept)s kept, %(self_total)s total services. "\ + "clustered new %(clustered_new)s, clustered vanished %(clustered_vanished)s" % result) + + # Note: Even if the actual mark-for-discovery flag may have been created by a cluster host, + # the activation decision is based on the discovery configuration of the node if redisc_params["activation"]: services_changed = True