diff options
Diffstat (limited to 'roles')
| -rw-r--r-- | roles/lib_zabbix/library/zbx_item.py | 2 | ||||
| -rw-r--r-- | roles/lib_zabbix/library/zbx_itemprototype.py | 6 | ||||
| -rw-r--r-- | roles/lib_zabbix/library/zbx_trigger.py | 24 | ||||
| -rw-r--r-- | roles/lib_zabbix/library/zbx_user_media.py | 3 | ||||
| -rwxr-xr-x | roles/openshift_facts/library/openshift_facts.py | 8 | ||||
| -rw-r--r-- | roles/openshift_manage_node/tasks/main.yml | 6 | ||||
| -rw-r--r-- | roles/openshift_node/templates/node.yaml.v1.j2 | 2 | ||||
| -rw-r--r-- | roles/os_zabbix/tasks/main.yml | 18 | ||||
| -rw-r--r-- | roles/os_zabbix/vars/template_app_zabbix_agent.yml | 23 | ||||
| -rw-r--r-- | roles/os_zabbix/vars/template_app_zabbix_server.yml | 408 | ||||
| -rw-r--r-- | roles/os_zabbix/vars/template_openshift_master.yml | 18 | ||||
| -rw-r--r-- | roles/os_zabbix/vars/template_openshift_node.yml | 28 | ||||
| -rw-r--r-- | roles/os_zabbix/vars/template_os_linux.yml | 40 |
13 files changed, 537 insertions, 49 deletions
diff --git a/roles/lib_zabbix/library/zbx_item.py b/roles/lib_zabbix/library/zbx_item.py index 2ccc21292..6faa82dfc 100644 --- a/roles/lib_zabbix/library/zbx_item.py +++ b/roles/lib_zabbix/library/zbx_item.py @@ -53,6 +53,8 @@ def get_value_type(value_type): vtype = 0 if 'int' in value_type: vtype = 3 + elif 'log' in value_type: + vtype = 2 elif 'char' in value_type: vtype = 1 elif 'str' in value_type: diff --git a/roles/lib_zabbix/library/zbx_itemprototype.py b/roles/lib_zabbix/library/zbx_itemprototype.py index 4ec1b8e02..e7fd6fa21 100644 --- a/roles/lib_zabbix/library/zbx_itemprototype.py +++ b/roles/lib_zabbix/library/zbx_itemprototype.py @@ -128,12 +128,12 @@ def get_status(status): return _status -def get_app_ids(zapi, application_names): +def get_app_ids(zapi, application_names, templateid): ''' get application ids from names ''' app_ids = [] for app_name in application_names: - content = zapi.get_content('application', 'get', {'search': {'name': app_name}}) + content = zapi.get_content('application', 'get', {'filter': {'name': app_name}, 'templateids': templateid}) if content.has_key('result'): app_ids.append(content['result'][0]['applicationid']) return app_ids @@ -212,7 +212,7 @@ def main(): 'ruleid': get_rule_id(zapi, module.params['discoveryrule_key'], template['templateid']), 'type': get_type(module.params['ztype']), 'value_type': get_value_type(module.params['value_type']), - 'applications': get_app_ids(zapi, module.params['applications']), + 'applications': get_app_ids(zapi, module.params['applications'], template['templateid']), 'description': module.params['description'], } diff --git a/roles/lib_zabbix/library/zbx_trigger.py b/roles/lib_zabbix/library/zbx_trigger.py index 21d0fcfd2..ab7731faa 100644 --- a/roles/lib_zabbix/library/zbx_trigger.py +++ b/roles/lib_zabbix/library/zbx_trigger.py @@ -86,6 +86,24 @@ def get_trigger_status(inc_status): return r_status +def get_template_id(zapi, template_name): + ''' + get related templates + ''' + template_ids = [] + app_ids = {} + # Fetch templates by name + content = zapi.get_content('template', + 'get', + {'search': {'host': template_name}, + 'selectApplications': ['applicationid', 'name']}) + if content.has_key('result'): + template_ids.append(content['result'][0]['templateid']) + for app in content['result'][0]['applications']: + app_ids[app['name']] = app['applicationid'] + + return template_ids, app_ids + def main(): ''' Create a trigger in zabbix @@ -117,6 +135,7 @@ def main(): url=dict(default=None, type='str'), status=dict(default=None, type='str'), state=dict(default='present', type='str'), + template_name=dict(default=None, type='str'), ), #supports_check_mode=True ) @@ -132,11 +151,16 @@ def main(): state = module.params['state'] tname = module.params['name'] + templateid = None + if module.params['template_name']: + templateid, _ = get_template_id(zapi, module.params['template_name']) + content = zapi.get_content(zbx_class_name, 'get', {'filter': {'description': tname}, 'expandExpression': True, 'selectDependencies': 'triggerid', + 'templateids': templateid, }) # Get diff --git a/roles/lib_zabbix/library/zbx_user_media.py b/roles/lib_zabbix/library/zbx_user_media.py index 9ed838f81..8895c78c3 100644 --- a/roles/lib_zabbix/library/zbx_user_media.py +++ b/roles/lib_zabbix/library/zbx_user_media.py @@ -260,6 +260,9 @@ def main(): for user in params['users']: diff['users']['userid'] = user['userid'] + # Medias have no real unique key so therefore we need to make it like the incoming user's request + diff['medias'] = medias + # We have differences and need to update content = zapi.get_content(zbx_class_name, 'updatemedia', diff) diff --git a/roles/openshift_facts/library/openshift_facts.py b/roles/openshift_facts/library/openshift_facts.py index 987f7f7da..69bb49c9b 100755 --- a/roles/openshift_facts/library/openshift_facts.py +++ b/roles/openshift_facts/library/openshift_facts.py @@ -480,11 +480,11 @@ def set_deployment_facts_if_unset(facts): if role in facts: deployment_type = facts['common']['deployment_type'] if 'registry_url' not in facts[role]: - registry_url = 'aos3/aos-${component}:${version}' - if deployment_type in ['enterprise', 'online']: + registry_url = 'openshift/origin-${component}:${version}' + if deployment_type in ['enterprise', 'online', 'openshift-enterprise']: registry_url = 'openshift3/ose-${component}:${version}' - elif deployment_type == 'origin': - registry_url = 'openshift/origin-${component}:${version}' + elif deployment_type == 'atomic-enterprise': + registry_url = 'aep3/aep-${component}:${version}' facts[role]['registry_url'] = registry_url return facts diff --git a/roles/openshift_manage_node/tasks/main.yml b/roles/openshift_manage_node/tasks/main.yml index 94d7879b2..637e494ea 100644 --- a/roles/openshift_manage_node/tasks/main.yml +++ b/roles/openshift_manage_node/tasks/main.yml @@ -1,6 +1,6 @@ - name: Wait for Node Registration command: > - {{ openshift.common.client_binary }} get node {{ item }} + {{ openshift.common.client_binary }} get node {{ item | lower }} register: omd_get_node until: omd_get_node.rc == 0 retries: 20 @@ -9,13 +9,13 @@ - name: Set node schedulability command: > - {{ openshift.common.admin_binary }} manage-node {{ item.openshift.common.hostname }} --schedulable={{ 'true' if item.openshift.node.schedulable | bool else 'false' }} + {{ openshift.common.admin_binary }} manage-node {{ item.openshift.common.hostname | lower }} --schedulable={{ 'true' if item.openshift.node.schedulable | bool else 'false' }} with_items: - "{{ openshift_node_vars }}" - name: Label nodes command: > - {{ openshift.common.client_binary }} label --overwrite node {{ item.openshift.common.hostname }} {{ item.openshift.node.labels | oo_combine_dict }} + {{ openshift.common.client_binary }} label --overwrite node {{ item.openshift.common.hostname | lower }} {{ item.openshift.node.labels | oo_combine_dict }} with_items: - "{{ openshift_node_vars }}" when: "'labels' in item.openshift.node and item.openshift.node.labels != {}" diff --git a/roles/openshift_node/templates/node.yaml.v1.j2 b/roles/openshift_node/templates/node.yaml.v1.j2 index 07d80f99b..946c0b655 100644 --- a/roles/openshift_node/templates/node.yaml.v1.j2 +++ b/roles/openshift_node/templates/node.yaml.v1.j2 @@ -18,7 +18,7 @@ networkPluginName: {{ openshift.common.sdn_network_plugin_name }} networkConfig: mtu: {{ openshift.node.sdn_mtu }} networkPluginName: {{ openshift.common.sdn_network_plugin_name }} -nodeName: {{ openshift.common.hostname }} +nodeName: {{ openshift.common.hostname | lower }} podManifestConfig: servingInfo: bindAddress: 0.0.0.0:10250 diff --git a/roles/os_zabbix/tasks/main.yml b/roles/os_zabbix/tasks/main.yml index 28e900255..a503b24d7 100644 --- a/roles/os_zabbix/tasks/main.yml +++ b/roles/os_zabbix/tasks/main.yml @@ -13,6 +13,8 @@ - include_vars: template_openshift_master.yml - include_vars: template_openshift_node.yml - include_vars: template_ops_tools.yml +- include_vars: template_app_zabbix_server.yml +- include_vars: template_app_zabbix_agent.yml - name: Include Template Heartbeat include: ../../lib_zabbix/tasks/create_template.yml @@ -61,3 +63,19 @@ server: "{{ ozb_server }}" user: "{{ ozb_user }}" password: "{{ ozb_password }}" + +- name: Include Template App Zabbix Server + include: ../../lib_zabbix/tasks/create_template.yml + vars: + template: "{{ g_template_app_zabbix_server }}" + server: "{{ ozb_server }}" + user: "{{ ozb_user }}" + password: "{{ ozb_password }}" + +- name: Include Template App Zabbix Agent + include: ../../lib_zabbix/tasks/create_template.yml + vars: + template: "{{ g_template_app_zabbix_agent }}" + server: "{{ ozb_server }}" + user: "{{ ozb_user }}" + password: "{{ ozb_password }}" diff --git a/roles/os_zabbix/vars/template_app_zabbix_agent.yml b/roles/os_zabbix/vars/template_app_zabbix_agent.yml new file mode 100644 index 000000000..06c4eda8b --- /dev/null +++ b/roles/os_zabbix/vars/template_app_zabbix_agent.yml @@ -0,0 +1,23 @@ +--- +g_template_app_zabbix_agent: + name: Template App Zabbix Agent + zitems: + - key: agent.hostname + applications: + - Zabbix agent + value_type: character + zabbix_type: '0' + + - key: agent.ping + applications: + - Zabbix agent + description: The agent always returns 1 for this item. It could be used in combination with nodata() for availability check. + value_type: int + zabbix_type: '0' + + ztriggers: + - name: '[Reboot] Zabbix agent on {HOST.NAME} is unreachable for 15 minutes' + description: Zabbix agent is unreachable for 15 minutes. + expression: '{Template App Zabbix Agent:agent.ping.nodata(15m)}=1' + priority: high + url: https://github.com/openshift/ops-sop/blob/master/Alerts/check_ping.asciidoc diff --git a/roles/os_zabbix/vars/template_app_zabbix_server.yml b/roles/os_zabbix/vars/template_app_zabbix_server.yml new file mode 100644 index 000000000..dace2aa29 --- /dev/null +++ b/roles/os_zabbix/vars/template_app_zabbix_server.yml @@ -0,0 +1,408 @@ +--- +g_template_app_zabbix_server: + name: Template App Zabbix Server + zitems: + - key: housekeeper_creates + applications: + - Zabbix server + description: A simple count of the number of partition creates output by the housekeeper script. + units: '' + value_type: int + zabbix_type: '2' + + - key: housekeeper_drops + applications: + - Zabbix server + description: A simple count of the number of partition drops output by the housekeeper script. + units: '' + value_type: int + zabbix_type: '2' + + - key: housekeeper_errors + applications: + - Zabbix server + description: A simple count of the number of errors output by the housekeeper script. + units: '' + value_type: int + zabbix_type: '2' + + - key: housekeeper_total + applications: + - Zabbix server + description: A simple count of the total number of lines output by the housekeeper + script. + units: '' + value_type: int + zabbix_type: '2' + + - key: zabbix[process,alerter,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,configuration syncer,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,db watchdog,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,discoverer,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,escalator,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,history syncer,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,housekeeper,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,http poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,icmp pinger,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,ipmi poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,java poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,node watcher,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,proxy poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,self-monitoring,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,snmp trapper,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,timer,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,trapper,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[process,unreachable poller,avg,busy] + applications: + - Zabbix server + description: '' + units: '%' + value_type: float + zabbix_type: '5' + + - key: zabbix[queue,10m] + applications: + - Zabbix server + description: '' + units: '' + value_type: int + zabbix_type: '5' + + - key: zabbix[queue] + applications: + - Zabbix server + description: '' + units: '' + value_type: int + zabbix_type: '5' + + - key: zabbix[rcache,buffer,pfree] + applications: + - Zabbix server + description: '' + units: '' + value_type: float + zabbix_type: '5' + + - key: zabbix[wcache,history,pfree] + applications: + - Zabbix server + description: '' + units: '' + value_type: float + zabbix_type: '5' + + - key: zabbix[wcache,text,pfree] + applications: + - Zabbix server + description: '' + units: '' + value_type: float + zabbix_type: '5' + + - key: zabbix[wcache,trend,pfree] + applications: + - Zabbix server + description: '' + units: '' + value_type: float + zabbix_type: '5' + + - key: zabbix[wcache,values] + applications: + - Zabbix server + description: '' + units: '' + value_type: float + zabbix_type: '5' + ztriggers: + - description: "There has been unexpected output while running the housekeeping script\ + \ on the Zabbix. There are only three kinds of lines we expect to see in the output,\ + \ and we've gotten something enw.\r\n\r\nCheck the script's output in /var/lib/zabbix/state\ + \ for more details." + expression: '{Template App Zabbix Server:housekeeper_errors.last(0)}+{Template App Zabbix Server:housekeeper_creates.last(0)}+{Template App Zabbix Server:housekeeper_drops.last(0)}<>{Template App Zabbix Server:housekeeper_total.last(0)}' + name: Unexpected output in Zabbix DB Housekeeping + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_DB_Housekeeping.asciidoc + + - description: An error has occurred during running the housekeeping script on the Zabbix. Check the script's output in /var/lib/zabbix/state for more details. + expression: '{Template App Zabbix Server:housekeeper_errors.last(0)}>0' + name: Errors during Zabbix DB Housekeeping + priority: high + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,alerter,avg,busy].min(600)}>75' + name: Zabbix alerter processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,configuration syncer,avg,busy].min(600)}>75' + name: Zabbix configuration syncer processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,db watchdog,avg,busy].min(600)}>75' + name: Zabbix db watchdog processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,discoverer,avg,busy].min(600)}>75' + name: Zabbix discoverer processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,escalator,avg,busy].min(600)}>75' + name: Zabbix escalator processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,history syncer,avg,busy].min(600)}>75' + name: Zabbix history syncer processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,housekeeper,avg,busy].min(1800)}>75' + name: Zabbix housekeeper processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,http poller,avg,busy].min(600)}>75' + name: Zabbix http poller processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,icmp pinger,avg,busy].min(600)}>75' + name: Zabbix icmp pinger processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,ipmi poller,avg,busy].min(600)}>75' + name: Zabbix ipmi poller processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,java poller,avg,busy].min(600)}>75' + name: Zabbix java poller processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,node watcher,avg,busy].min(600)}>75' + name: Zabbix node watcher processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,poller,avg,busy].min(600)}>75' + name: Zabbix poller processes more than 75% busy + priority: high + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,proxy poller,avg,busy].min(600)}>75' + name: Zabbix proxy poller processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,self-monitoring,avg,busy].min(600)}>75' + name: Zabbix self-monitoring processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,snmp trapper,avg,busy].min(600)}>75' + name: Zabbix snmp trapper processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: Timer processes usually are busy because they have to process time + based trigger functions + expression: '{Template App Zabbix Server:zabbix[process,timer,avg,busy].min(600)}>75' + name: Zabbix timer processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,trapper,avg,busy].min(600)}>75' + name: Zabbix trapper processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[process,unreachable poller,avg,busy].min(600)}>75' + name: Zabbix unreachable poller processes more than 75% busy + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/Zabbix_state_check.asciidoc + + - description: "This alert generally indicates a performance problem or a problem\ + \ with the zabbix-server or proxy.\r\n\r\nThe first place to check for issues\ + \ is Administration > Queue. Be sure to check the general view and the per-proxy\ + \ view." + expression: '{Template App Zabbix Server:zabbix[queue,10m].min(600)}>1000' + name: More than 1000 items having missing data for more than 10 minutes + priority: high + url: https://github.com/openshift/ops-sop/blob/master/Alerts/data_lost_overview_plugin.asciidoc + + - description: Consider increasing CacheSize in the zabbix_server.conf configuration + file + expression: '{Template App Zabbix Server:zabbix[rcache,buffer,pfree].min(600)}<5' + name: Less than 5% free in the configuration cache + priority: info + url: https://github.com/openshift/ops-sop/blob/master/Alerts/check_cache.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[wcache,history,pfree].min(600)}<25' + name: Less than 25% free in the history cache + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/check_cache.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[wcache,text,pfree].min(600)}<25' + name: Less than 25% free in the text history cache + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/check_cache.asciidoc + + - description: '' + expression: '{Template App Zabbix Server:zabbix[wcache,trend,pfree].min(600)}<25' + name: Less than 25% free in the trends cache + priority: avg + url: https://github.com/openshift/ops-sop/blob/master/Alerts/check_cache.asciidoc diff --git a/roles/os_zabbix/vars/template_openshift_master.yml b/roles/os_zabbix/vars/template_openshift_master.yml index c71e07910..4ae918ec6 100644 --- a/roles/os_zabbix/vars/template_openshift_master.yml +++ b/roles/os_zabbix/vars/template_openshift_master.yml @@ -13,6 +13,18 @@ g_template_openshift_master: applications: - Openshift Master + - key: openshift.master.user.count + description: Shows number of users in a cluster + type: int + applications: + - Openshift Master + + - key: openshift.master.pod.running.count + description: Shows number of pods running + type: int + applications: + - Openshift Master + ztriggers: - name: 'Application creation has failed on {HOST.NAME}' expression: '{Template Openshift Master:create_app.last(#1)}=1 and {Template Openshift Master:create_app.last(#2)}=1' @@ -28,3 +40,9 @@ g_template_openshift_master: expression: '{Template Openshift Master:openshift.master.process.count.min(#3)}>1' url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/openshift_master.asciidoc' priority: high + + - name: 'Number of users for Openshift Master on {HOST.NAME}' + expression: '{Template Openshift Master:openshift.master.user.count.last()}=0' + url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/openshift_master.asciidoc' + priority: info + diff --git a/roles/os_zabbix/vars/template_openshift_node.yml b/roles/os_zabbix/vars/template_openshift_node.yml index 36f9cc4a3..ce28b1048 100644 --- a/roles/os_zabbix/vars/template_openshift_node.yml +++ b/roles/os_zabbix/vars/template_openshift_node.yml @@ -8,13 +8,37 @@ g_template_openshift_node: applications: - Openshift Node + - key: openshift.node.ovs.pids.count + description: Shows number of ovs process ids running + type: int + applications: + - Openshift Node + + - key: openshift.node.ovs.ports.count + description: Shows number of OVS ports defined + type: int + applications: + - Openshift Node + ztriggers: - name: 'Openshift Node process not running on {HOST.NAME}' expression: '{Template Openshift Node:openshift.node.process.count.max(#3)}<1' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/openshift_node.asciidoc' + url: 'https://github.com/openshift/ops-sop/blob/node/V3/Alerts/openshift_node.asciidoc' priority: high - name: 'Too many Openshift Node processes running on {HOST.NAME}' expression: '{Template Openshift Node:openshift.node.process.count.min(#3)}>1' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/openshift_node.asciidoc' + url: 'https://github.com/openshift/ops-sop/blob/node/V3/Alerts/openshift_node.asciidoc' + priority: high + + - name: 'OVS may not be running on {HOST.NAME}' + expression: '{Template Openshift Node:openshift.node.ovs.pids.count.last()}<>4' + url: 'https://github.com/openshift/ops-sop/blob/node/V3/Alerts/openshift_node.asciidoc' priority: high + + - name: 'Number of OVS ports is 0 on {HOST.NAME}' + expression: '{Template Openshift Node:openshift.node.ovs.ports.count.last()}=0' + url: 'https://github.com/openshift/ops-sop/blob/node/V3/Alerts/openshift_node.asciidoc' + priority: high + + diff --git a/roles/os_zabbix/vars/template_os_linux.yml b/roles/os_zabbix/vars/template_os_linux.yml index 70c3809bd..cd9649773 100644 --- a/roles/os_zabbix/vars/template_os_linux.yml +++ b/roles/os_zabbix/vars/template_os_linux.yml @@ -188,18 +188,6 @@ g_template_os_linux: multiplier: 1024 units: B - # Disk items - - key: filesys.full.xvda2 - applications: - - Disk - value_type: float - - - key: filesys.full.xvda3 - applications: - - Disk - value_type: float - - zdiscoveryrules: - name: disc.filesys key: disc.filesys @@ -216,37 +204,17 @@ g_template_os_linux: - Disk ztriggerprototypes: - - name: 'Filesystem: {#OSO_FILESYS} has less than 10% free on {HOST.NAME}' - expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>90' + - name: 'Filesystem: {#OSO_FILESYS} has less than 15% free on {HOST.NAME}' + expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>85' url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' priority: warn - - name: 'Filesystem: {#OSO_FILESYS} has less than 5% free on {HOST.NAME}' - expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>95' + - name: 'Filesystem: {#OSO_FILESYS} has less than 10% free on {HOST.NAME}' + expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>90' url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' priority: high ztriggers: - - name: 'Filesystem: / has less than 10% free on {HOST.NAME}' - expression: '{Template OS Linux:filesys.full.xvda2.last()}>90' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' - priority: warn - - - name: 'Filesystem: / has less than 5% free on {HOST.NAME}' - expression: '{Template OS Linux:filesys.full.xvda2.last()}>95' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' - priority: high - - - name: 'Filesystem: /var has less than 10% free on {HOST.NAME}' - expression: '{Template OS Linux:filesys.full.xvda3.last()}>90' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' - priority: warn - - - name: 'Filesystem: /var has less than 5% free on {HOST.NAME}' - expression: '{Template OS Linux:filesys.full.xvda3.last()}>95' - url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc' - priority: high - - name: 'Too many TOTAL processes on {HOST.NAME}' expression: '{Template OS Linux:proc.nprocs.last()}>5000' url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_proc.asciidoc' |