cyberes
/
icinga2-checks
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

also process hosts

This commit is contained in:
Cyberes 2023-11-22 23:51:23 -07:00
parent b6149a7918
commit e3664a66cc
1 changed files with 42 additions and 33 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

75
fail-overdue.py
View File

@ -19,44 +19,53 @@ def main(args):
current_time = time.time() current_time = time.time()
# Get all checks # Get all checks
response = requests.get(url + "/objects/services", auth=(args.username, args.password), verify=not args.insecure) objects = ["hosts", "services"]
response.raise_for_status() for object in objects:
response = requests.get(url + "/objects/" + object, auth=(args.username, args.password), verify=not args.insecure)
response.raise_for_status()
checks = response.json()["results"] checks = response.json()["results"]
# Loop through all checks # Loop through all checks
for check in checks: for check in checks:
last_check_time = check["attrs"]["last_check"] last_check_time = check["attrs"]["last_check"]
check_interval = check["attrs"]["check_interval"] check_interval = check["attrs"]["check_interval"]
check_timeout = check["attrs"]["check_timeout"] or args.default_timeout check_timeout = check["attrs"]["check_timeout"] or args.default_timeout
# If the check is overdue. # If the check is overdue.
if current_time - last_check_time > check_interval + check_timeout and check['attrs']['state'] != 3: if current_time - last_check_time > check_interval + check_timeout and check['attrs']['state'] != 3:
headers = {'Accept': 'application/json', 'Content-Type': 'application/json'} headers = {'Accept': 'application/json', 'Content-Type': 'application/json'}
next_check_time = last_check_time + check_interval + check_timeout next_check_time = last_check_time + check_interval + check_timeout
check_filter = f'host.name=="{check["attrs"]["host_name"]}" && service.name=="{check["attrs"]["name"]}"'
# Set the check to unknown
data = {
"type": check['type'],
"filter": check_filter,
"exit_status": 3,
"plugin_output": f"<Check is overdue for {int(current_time - next_check_time)} seconds.>",
}
# Trigger a full failure. if object == 'services':
for _ in range(4): check_filter = f'host.name=="{check["attrs"]["host_name"]}" && service.name=="{check["attrs"]["name"]}"'
response = requests.post(url + "/actions/process-check-result", data=json.dumps(data), headers=headers, auth=(args.username, args.password), verify=not args.insecure) elif object == 'hosts':
check_filter = f'host.name=="{check["name"]}"'
else:
raise Exception
# Set the check to unknown
data = {
"type": check['type'],
"filter": check_filter,
"exit_status": 3 if check['type'] == 'Service' else 1,
"plugin_output": f"<Check is overdue for {int(current_time - next_check_time)} seconds.>",
}
# Trigger a full failure.
for _ in range(4):
response = requests.post(url + "/actions/process-check-result", data=json.dumps(data), headers=headers, auth=(args.username, args.password), verify=not args.insecure)
response.raise_for_status()
time.sleep(3)
# Rerun the check
data = {
"type": check['type'],
"filter": check_filter,
}
response = requests.post(url + "/actions/reschedule-check", data=json.dumps(data), headers=headers, auth=(args.username, args.password), verify=not args.insecure)
response.raise_for_status() response.raise_for_status()
time.sleep(3) logging.info(f'Failed {check["type"].lower()} {check["name"]} - {int(current_time - next_check_time)} seconds overdue.')
# Rerun the check
data = {
"type": check['type'],
"filter": check_filter,
}
response = requests.post(url + "/actions/reschedule-check", data=json.dumps(data), headers=headers, auth=(args.username, args.password), verify=not args.insecure)
response.raise_for_status()
logging.info(f'Failed {check["name"]} - {int(current_time - next_check_time)} seconds overdue.')
time.sleep(args.interval) time.sleep(args.interval)