ansible/roles/nagios_client/files/scripts/check_datanommer_timesince.py

75 lines
2.2 KiB
Python
Raw Normal View History

2020-10-05 15:17:25 +01:00
#!/usr/bin/python
""" NRPE check for datanommer/fedmsg health.
Given a category like 'bodhi', 'buildsys', or 'git', return an error if
datanommer hasn't seen a message of that type in such and such time.
You can alternatively provide a 'topic' which might look like
org.fedoraproject.prod.bodhi.update.comment.
2020-10-05 15:17:25 +01:00
Requires: python-dateutil
2020-10-05 15:17:25 +01:00
Usage:
2020-10-05 15:17:25 +01:00
$ check_datanommer_timesince CATEGORY WARNING_THRESH CRITICAL_THRESH
2020-10-05 15:17:25 +01:00
:Author: Ralph Bean <rbean@redhat.com>
2020-10-05 15:17:25 +01:00
"""
2020-10-05 15:17:25 +01:00
from __future__ import print_function
from builtins import str
import dateutil.relativedelta
import subprocess
import sys
import json
2020-10-05 15:17:25 +01:00
def query_timesince(identifier):
# If it has a '.', then assume it is a topic.
if '.' in identifier:
cmd = 'datanommer-latest --topic %s --timesince' % identifier
else:
cmd = 'datanommer-latest --category %s --timesince' % identifier
sys.stderr.write("Running %r\n" % cmd)
process = subprocess.Popen(cmd.split(), shell=False,
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
stdout, stderr = process.communicate()
prefix, stdout = stdout.split("INFO] ", 1)
data = json.loads(stdout)
return float(data[0])
2020-10-05 15:17:25 +01:00
def main():
identifier, warning_threshold, critical_threshold = sys.argv[-3:]
timesince = query_timesince(identifier)
warning_threshold = int(warning_threshold)
critical_threshold = int(critical_threshold)
2020-10-05 15:17:25 +01:00
time_strings = []
rd = dateutil.relativedelta.relativedelta(seconds=timesince)
for denomination in ['years', 'months', 'days', 'hours', 'minutes', 'seconds']:
value = getattr(rd, denomination, 0)
if value:
time_strings.append("%d %s" % (value, denomination))
2020-10-05 15:17:25 +01:00
string = ", ".join(time_strings)
reason = "datanommer has not seen a %r message in %s" % (identifier, string)
2020-10-05 15:17:25 +01:00
if timesince > critical_threshold:
2020-10-05 15:17:25 +01:00
print("CRIT: ", reason)
sys.exit(2)
2020-10-05 15:17:25 +01:00
if timesince > warning_threshold:
2020-10-05 15:17:25 +01:00
print("WARN: ", reason)
sys.exit(1)
2020-10-05 15:17:25 +01:00
print("OK: ", reason)
sys.exit(0)
2020-10-05 15:17:25 +01:00
if __name__ == '__main__':
try:
main()
except Exception as e:
2020-10-05 15:17:25 +01:00
print("UNKNOWN: ", str(e))
sys.exit(3)