# NOTE: there is a narrow window in which this command must be
# evaluated, otherwise the notice will not go out.
# this is not ideal.
- sitehist.sendMessage('online_notice', hostname=host, viart=False)
+ sitehist.sendMessage('online_notice', hostname=host, viart=False, saveact=True)
print "send message for host %s online" % host
# sitehist.sendMessage('retry_bootman', hostname=host)
if nodehist.status == 'down' and \
- changed_greaterthan(nodehist.last_changed, 2) and \
- not found_within(recent_actions, 'down_notice', 3.5):
- # send down node notice
-
- sitehist.sendMessage('down_notice', hostname=host)
- print "send message for host %s down" % host
+ changed_greaterthan(nodehist.last_changed, 2):
+ if not nodehist.firewall and not found_within(recent_actions, 'down_notice', 3.5):
+ # send down node notice
+ sitehist.sendMessage('down_notice', hostname=host)
+ print "send message for host %s down" % host
+
+ if nodehist.firewall and not found_within(recent_actions, 'firewall_notice', 3.5):
+ # send down node notice
+ email_exception(host, "firewall_notice")
+ sitehist.sendMessage('firewall_notice', hostname=host)
+ print "send message for host %s down" % host
node_count = node_count + 1
print "time: ", time.strftime('%Y-%m-%d %H:%M:%S')
for i,site in enumerate(sitenames):
sitehist = SiteInterface.get_or_make(loginbase=site)
siteblack = BlacklistRecord.get_by(loginbase=site)
+ skip_due_to_blacklist=False
if siteblack and not siteblack.expired():
print "skipping %s due to blacklist. will expire %s" % (site, siteblack.willExpire() )
+ skip_due_to_blacklist=True
+ sitehist.clearPenalty()
+ sitehist.applyPenalty()
continue
# TODO: make query only return records within a certin time range,
session.flush()
sys.exit(0)
except:
- #email_exception()
+ email_exception()
print traceback.print_exc();
print "fail all..."