entrouvert
/
publik-infra
16
0
Fork 0
Code Issues Pull Requests Releases Activity

journald: rework logic to respect ignored error (#78679)
gitea/publik-infra/pipeline/head This commit looks good Details

This commit is contained in:
Guillaume Baffoin 2023-08-22 10:35:34 +02:00
parent 6e39c3bce9
commit cf566de881
1 changed files with 12 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
prometheus-entrouvert-exporter/prometheus-system-exporter.py
View File

@ -27,6 +27,9 @@ JOURNALD_IGNORED_ERRORS = {
],
'ssh.service': [
'maximum authentication attempts exceeded for ',
'error: kex_exchange_identification: read: Connection reset by peer',
'error: kex_exchange_identification: Connection closed by remote host',
'fatal: Timeout before authentication for ',
],
'': [ # match all services (useful for ovpn*)
'Connection reset, restarting [0]',
@ -109,30 +112,27 @@ def harakiri(ctn):
if 'HARAKIRI ON WORKER' in e['MESSAGE']:
eo_harakiri.labels(ctn, 'errors', e['_SYSTEMD_UNIT']).inc()
def journald(ctn):
j = journal.Reader()
fifteen = time.time() - 15 * 60
j.seek_realtime(fifteen)
j.add_match(PRIORITY=2)
eo_journal.labels(ctn, 'critical').set(len(list(j)))
j.seek_realtime(fifteen)
j.add_match(PRIORITY=3)
j.add_match(PRIORITY=2)
eo_journal.labels(ctn, "critical").set(0)
eo_journal.labels(ctn, "error").set(0)
priority_to_tag = {2: "critical", 3: "error"}
for e in j:
msg = e['MESSAGE']
ignored_message = False
msg = e["MESSAGE"]
ignored_strings = (
JOURNALD_IGNORED_ERRORS.get(e.get('_SYSTEMD_UNIT'), []) + JOURNALD_IGNORED_ERRORS['']
)
for ignored_string in ignored_strings:
if ignored_string in msg:
ignored_message = True
break
if ignored_message:
continue
eo_journal.labels(ctn, 'error').inc()
if 'Connected -> NetworkFailure' in msg or 'task nfsd' in msg:
eo_journal.labels(ctn, 'network_failure').inc()
else:
eo_journal.labels(ctn, priority_to_tag[e["PRIORITY"]]).inc()
if "Connected -> NetworkFailure" in msg or "task nfsd" in msg:
eo_journal.labels(ctn, "network_failure").inc()
def local_changes(ctn):