Modify tempmon server logic to take "unfair" time windows into account

when a client or probe first are (re-)enabled, we can't expect to have readings
within the time window we'd normally be checking.  previously we'd get false
alarms about "probe error status" etc. when this happened; hopefully no longer!
This commit is contained in:
Lance Edgar 2019-01-25 19:49:46 -06:00
parent f31a0c4c22
commit cf27af81d4

View file

@ -122,10 +122,24 @@ class TempmonServerDaemon(Daemon):
# the client to be (possibly) offline.
delay = client.delay or 60
cutoff = self.now - datetime.timedelta(seconds=delay + 60)
# but if client was "just now" enabled, cutoff may not be quite fair.
# in this case we'll just skip checks until cutoff does seem fair.
if cutoff < client.enabled:
return
# we make similar checks for each probe; if cutoff "is not fair" for
# any of them, we'll skip that probe check, and avoid marking client
# offline for this round, just to be safe
online = False
cutoff_unfair = False
for probe in client.enabled_probes():
if self.check_readings_for_probe(session, probe, cutoff):
if cutoff < probe.enabled:
cutoff_unfair = True
elif self.check_readings_for_probe(session, probe, cutoff):
online = True
if cutoff_unfair:
return
# if client was previously marked online, but we have no "new"
# readings, then let's look closer to see if it's been long enough to