starlingx
/
integ
Code Issues Proposed changes

Merge "Remove alarm query before clear in NTP plugin"

This commit is contained in:
Zuul 2019-01-21 14:59:56 +00:00 committed by Gerrit Code Review
parent 7bb43963d3 abaff6b275
commit e12b3a436f
2 changed files with 21 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
monitoring/collectd-extensions/centos/build_srpm.data
View File

@ -16,4 +16,4 @@ COPY_LIST="$PKG_BASE/src/LICENSE \
$PKG_BASE/src/example.py \
$PKG_BASE/src/example.conf"
TIS_PATCH_VER=5
TIS_PATCH_VER=6

44
monitoring/collectd-extensions/src/ntpq.py
View File

@ -222,15 +222,14 @@ def _raise_alarm(ip=None):
def _clear_base_alarm():
""" Clear the NTP base alarm """
if api.get_fault(PLUGIN_ALARMID, obj.base_eid) is not None:
if api.clear_fault(PLUGIN_ALARMID, obj.base_eid) is False:
collectd.error("%s failed to clear alarm %s:%s" %
(PLUGIN, PLUGIN_ALARMID, obj.base_eid))
return True
else:
collectd.info("%s cleared alarm %s:%s" %
(PLUGIN, PLUGIN_ALARMID, obj.base_eid))
obj.alarm_raised = False
if api.clear_fault(PLUGIN_ALARMID, obj.base_eid) is False:
collectd.error("%s failed to clear alarm %s:%s" %
(PLUGIN, PLUGIN_ALARMID, obj.base_eid))
return True
else:
collectd.info("%s cleared alarm %s:%s" %
(PLUGIN, PLUGIN_ALARMID, obj.base_eid))
obj.alarm_raised = False
return False
@ -263,23 +262,20 @@ def _remove_ip_from_unreachable_list(ip):
if ip and ip in obj.unreachable_servers:
eid = obj.base_eid + '=' + ip
collectd.debug("%s trying to clear alarm %s" % (PLUGIN, eid))
# clear the alarm if its asserted
if api.get_fault(PLUGIN_ALARMID, eid) is not None:
if api.clear_fault(PLUGIN_ALARMID, eid) is True:
collectd.info("%s cleared %s:%s alarm" %
(PLUGIN, PLUGIN_ALARMID, eid))
obj.unreachable_servers.remove(ip)
else:
# Handle clear failure by not removing the IP from the list.
# It will retry on next audit.
# Error should only occur if FM is not running at the time
# this get or clear is called
collectd.error("%s failed alarm clear %s:%s" %
(PLUGIN, PLUGIN_ALARMID, eid))
return True
else:
if api.clear_fault(PLUGIN_ALARMID, eid) is True:
collectd.info("%s cleared %s:%s alarm" %
(PLUGIN, PLUGIN_ALARMID, eid))
obj.unreachable_servers.remove(ip)
collectd.info("%s alarm %s not raised" % (PLUGIN, eid))
else:
# Handle clear failure by not removing the IP from the list.
# It will retry on next audit.
# Error should only occur if FM is not running at the time
# this get or clear is called
collectd.error("%s failed alarm clear %s:%s" %
(PLUGIN, PLUGIN_ALARMID, eid))
return True
return False