def resolve_patterns(request, expression_values):
now = int(time())
context = createRequestContext(str(now - 600),
str(now),
allowRealTimeAlerting=True)
resolved = set()
target_num = 1
context['time_series_names'] = set()
is_simple_trigger = True
if len(request.body_json["targets"]) > 1:
is_simple_trigger = False
for target in request.body_json["targets"]:
time_series = yield evaluateTarget(context, target)
if is_simple_trigger and not is_simple_target(context):
is_simple_trigger = False
target_name = "t%s" % target_num
for ts in time_series:
context['time_series_names'].add(ts.name)
expression_values[target_name] = 42
target_num += 1
for pattern, resolve in context['graphite_patterns'].iteritems():
for r in resolve:
if r != pattern:
resolved.add(r)
request.body_json["patterns"] = [
pattern for pattern in context['graphite_patterns']
if pattern not in resolved
]
request.body_json["is_simple_trigger"] = is_simple_trigger
request.context = context
def resolve_patterns(request, expression_values):
now = int(time())
context = createRequestContext(str(now - 600), str(now), allowRealTimeAlerting=True)
resolved = set()
target_num = 1
context['time_series_names'] = set()
is_simple_trigger = True
if len(request.body_json["targets"]) > 1:
is_simple_trigger = False
for target in request.body_json["targets"]:
time_series = yield evaluateTarget(context, target)
if is_simple_trigger and not is_simple_target(context):
is_simple_trigger = False
target_name = "t%s" % target_num
for ts in time_series:
context['time_series_names'].add(ts.name)
expression_values[target_name] = 42
target_num += 1
for pattern, resolve in context['graphite_patterns'].iteritems():
for r in resolve:
if r != pattern:
resolved.add(r)
request.body_json["patterns"] = [pattern for pattern in context['graphite_patterns']
if pattern not in resolved]
request.body_json["is_simple_trigger"] = is_simple_trigger
request.context = context
def movingAverageBootstrap(self, allowRealTimeAlerting):
yield self.sendTrigger(
'{"name": "t", "targets": ["movingAverage(m, 2)"], "warn_value": 1, "error_value": 90, "ttl":"600" }')
for n in range(0, 10):
yield self.db.sendMetric('m', 'm', self.now - 60 * (10 - n), n)
yield self.trigger.check(fromTime=self.now - 300, now=self.now)
fromTime = str(self.now - 180)
endTime = str(self.now - 60)
rc = datalib.createRequestContext(fromTime, endTime, allowRealTimeAlerting)
result = yield self.trigger.get_timeseries(rc)
ts = result[1][0]
self.assertEqual(ts[0], 6.5)
def movingAverageBootstrap(self, allowRealTimeAlerting):
yield self.sendTrigger(
'{"name": "t", "targets": ["movingAverage(m, 2)"], "warn_value": 1, "error_value": 90, "ttl":"600" }'
)
for n in range(0, 10):
yield self.db.sendMetric('m', 'm', self.now - 60 * (10 - n), n)
yield self.trigger.check(fromTime=self.now - 300, now=self.now)
fromTime = str(self.now - 180)
endTime = str(self.now - 60)
rc = datalib.createRequestContext(fromTime, endTime,
allowRealTimeAlerting)
result = yield self.trigger.get_timeseries(rc)
ts = result[1][0]
self.assertEqual(ts[0], 6.5)
def render_GET(self, request):
json, trigger = yield self.db.getTrigger(self.trigger_id)
if json is None:
defer.returnValue(bad_request(request, "Trigger not found"))
raise StopIteration
context = createRequestContext(request.args.get('from')[0],
request.args.get('to')[0])
result = {}
for target in trigger.get("targets", [trigger.get("target")]):
time_series = yield evaluateTarget(context, target)
for time_serie in time_series:
values = [(time_serie.start + time_serie.step * i, time_serie[i]) for i in range(0, len(time_serie))]
result[time_serie.name] = [{"ts": ts, "value": value} for ts, value in values if value is not None]
self.write_json(request, result)
def render_GET(self, request):
json, trigger = yield self.db.getTrigger(self.trigger_id)
if json is None:
defer.returnValue(bad_request(request, "Trigger not found"))
raise StopIteration
fromTime = request.args.get('from')[0]
endTime = request.args.get('to')[0]
context = createRequestContext(fromTime, endTime, allowRealTimeAlerting=True)
result = {}
for target in trigger.get("targets", [trigger.get("target")]):
time_series = yield evaluateTarget(context, target)
for time_serie in time_series:
values = [(time_serie.start + time_serie.step * i, time_serie[i]) for i in range(0, len(time_serie))]
result[time_serie.name] = [{"ts": ts, "value": value} for ts, value in values if value is not None]
self.write_json(request, result)
def resolve_patterns(request, expression_values):
now = int(time())
context = createRequestContext(str(now - 10), str(now))
resolved = set()
target_num = 1
for target in request.body_json["targets"]:
yield evaluateTarget(context, target)
target_name = "t%s" % target_num
expression_values[target_name] = 42
target_num += 1
for pattern, resolve in context['graphite_patterns'].iteritems():
for r in resolve:
if r != pattern:
resolved.add(r)
request.body_json["patterns"] = [pattern for pattern in context['graphite_patterns']
if pattern not in resolved]
def resolve_patterns(request, expression_values):
now = int(time())
requestContext = createRequestContext(str(now - 10), str(now))
resolved = set()
target_num = 1
for target in request.body_json["targets"]:
target_time_series = yield evaluateTarget(requestContext, target)
target_name = "t%s" % target_num
expression_values[target_name] = 42
if len(target_time_series) > 0 and len(target_time_series[0]) > 0:
expression_values[target_name] = target_time_series[0][-1]
target_num += 1
for pattern, resolve in requestContext['graphite_patterns'].iteritems():
for r in resolve:
if r != pattern:
resolved.add(r)
request.body_json["patterns"] = [pattern for pattern in requestContext['graphite_patterns']
if pattern not in resolved]
def migrate_triggers(db):
now = int(time())
trigger_ids = yield db.getTriggers()
logs.log.info("triggers count: %d" % len(trigger_ids))
converted_triggers_count = 0
simple_triggers_count = 0
complex_triggers_count = 0
failed_triggers_count = 0
for trigger_id in trigger_ids:
try:
json, _ = yield db.getTrigger(trigger_id)
if json is None:
continue
trigger = anyjson.deserialize(json)
if "is_simple_trigger" in trigger:
continue
logs.log.info("recalculating for trigger %s (%s)" %
(trigger_id, trigger.get("name")))
context = createRequestContext(str(now - 600),
str(now),
allowRealTimeAlerting=True)
if len(trigger["targets"]) != 1:
is_simple_trigger = False
else:
yield evaluateTarget(context, trigger["targets"][0])
is_simple_trigger = is_simple_target(context)
trigger["is_simple_trigger"] = is_simple_trigger
logs.log.info(str(trigger["is_simple_trigger"]))
yield db.saveTrigger(trigger_id, trigger)
converted_triggers_count += 1
if is_simple_trigger:
simple_triggers_count += 1
else:
complex_triggers_count += 1
except Exception, e:
failed_triggers_count += 1
logs.log.error("conversion failed for trigger: %s" % e)
def trigger(trigger, fromTime, now, cache_ttl):
now = now or int(time())
log.info("Checking trigger {id}", id=trigger.id)
initialized = yield trigger.init(now, fromTime=fromTime)
if not initialized:
raise StopIteration
if fromTime is None:
fromTime = trigger.last_check.get("timestamp", now)
requestContext = datalib.createRequestContext(
str(fromTime - (trigger.ttl or 600)), str(now))
check = {
"metrics": trigger.last_check["metrics"].copy(),
"state": state.OK,
"timestamp": now,
"score": trigger.last_check.get("score")
}
try:
time_series = yield trigger.get_timeseries(requestContext)
for metric in requestContext['metrics']:
yield trigger.db.cleanupMetricValues(metric,
now - config.METRICS_TTL,
cache_key=metric,
cache_ttl=cache_ttl)
if not time_series:
if trigger.ttl:
check["state"] = trigger.ttl_state
check["msg"] = "Trigger has no metrics"
yield event.compare_states(trigger, check, trigger.last_check,
now)
else:
for t_series in time_series.values():
for tN in t_series:
if not tN.stub:
check["metrics"][tN.name] = tN.last_state.copy()
for t1 in time_series[1]:
log.debug("Checking timeserie {name}: {values}",
name=t1.name,
values=list(t1))
log.debug(
"Checking interval: {start} - {end} ({duration}s), step: {step}",
start=t1.start,
end=t1.end,
step=t1.step,
duration=t1.end - t1.start)
metric_state = check["metrics"].get(t1.name)
if not metric_state:
log.debug("No metric state for {name}.", name=t1.name)
continue
checkpoint = max(
t1.last_state["timestamp"] - config.CHECKPOINT_GAP,
metric_state.get("event_timestamp", 0))
log.debug("Checkpoint for {name}: {checkpoint}",
name=t1.name,
checkpoint=checkpoint)
for value_timestamp in xrange(t1.start, now + t1.step,
t1.step):
if value_timestamp <= checkpoint:
continue
expression_values = time_series.get_expression_values(
t1, value_timestamp)
t1_value = expression_values["t1"]
log.debug("values for ts {timestamp}: {values}",
timestamp=value_timestamp,
values=expression_values)
if None in expression_values.values():
continue
expression_values.update({
'warn_value':
trigger.struct.get('warn_value'),
'error_value':
trigger.struct.get('error_value'),
'PREV_STATE':
metric_state['state']
})
expression_state = expression.getExpression(
trigger.struct.get('expression'), **expression_values)
time_series.update_state(t1, check, expression_state,
expression_values,
value_timestamp)
yield event.compare_states(trigger,
metric_state,
t1.last_state,
value_timestamp,
value=t1_value,
metric=t1.name)
# compare with last_check timestamp in case if we have not run checker for a long time
if trigger.ttl and metric_state[
"timestamp"] + trigger.ttl < trigger.last_check[
"timestamp"]:
log.info("Metric {name} TTL expired for state {state}",
name=t1.name,
state=metric_state)
if trigger.ttl_state == state.DEL and metric_state.get(
"event_timestamp") is not None:
log.info("Remove metric {name}", name=t1.name)
del check["metrics"][t1.name]
for tN, tName in time_series.other_targets_names.iteritems(
):
log.info("Remove metric {name}", name=tName)
del check["metrics"][tName]
for pattern in trigger.struct.get("patterns"):
yield trigger.db.delPatternMetrics(pattern)
continue
time_series.update_state(
t1, check, state.to_metric_state(trigger.ttl_state),
None, trigger.last_check["timestamp"] - trigger.ttl)
yield event.compare_states(trigger,
metric_state,
t1.last_state,
metric_state["timestamp"],
metric=t1.name)
except StopIteration:
raise
except Exception as e:
log.error("Trigger check failed: {e}", e=e)
check["state"] = state.EXCEPTION
check["msg"] = "Trigger evaluation exception"
yield event.compare_states(trigger, check, trigger.last_check, now)
scores = sum(
map(lambda m: state.SCORES[m["state"]], check["metrics"].itervalues()))
check["score"] = scores + state.SCORES[check["state"]]
yield trigger.db.setTriggerLastCheck(trigger.id, check)
def trigger(trigger, fromTime, now, cache_ttl):
now = now or int(time())
log.msg("Checking trigger %s" % trigger.id)
initialized = yield trigger.init(now, fromTime=fromTime)
if not initialized:
raise StopIteration
if fromTime is None:
fromTime = trigger.last_check.get("timestamp", now)
requestContext = datalib.createRequestContext(
str(fromTime - (trigger.ttl or 600)), str(now))
check = {
"metrics": trigger.last_check["metrics"].copy(),
"state": state.OK,
"timestamp": now,
"score": trigger.last_check.get("score")
}
try:
time_series = yield trigger.get_timeseries(requestContext)
for metric in requestContext['metrics']:
yield trigger.db.cleanupMetricValues(metric,
now - config.METRICS_TTL,
cache_key=metric,
cache_ttl=cache_ttl)
if not time_series:
if trigger.ttl:
check["state"] = trigger.ttl_state
check["msg"] = "Trigger has no metrics"
yield event.compare_states(trigger, check, trigger.last_check,
now)
else:
for t_series in time_series.values():
for tN in t_series:
if not tN.stub:
check["metrics"][tN.name] = tN.last_state.copy()
for t1 in time_series[1]:
metric_state = check["metrics"].get(t1.name)
if not metric_state:
continue
checkpoint = max(
t1.last_state["timestamp"] - config.CHECKPOINT_GAP,
metric_state.get("event_timestamp", 0))
for value_timestamp in xrange(t1.start, now + t1.step,
t1.step):
if value_timestamp <= checkpoint:
continue
expression_values = time_series.get_expression_values(
t1, value_timestamp)
t1_value = expression_values["t1"]
if None in expression_values.values():
continue
expression_values.update({
'warn_value':
trigger.struct.get('warn_value'),
'error_value':
trigger.struct.get('error_value'),
'PREV_STATE':
metric_state['state']
})
expression_state = expression.getExpression(
trigger.struct.get('expression'), **expression_values)
time_series.update_state(t1, check, expression_state,
expression_values,
value_timestamp)
yield event.compare_states(trigger,
metric_state,
t1.last_state,
value_timestamp,
value=t1_value,
metric=t1.name)
# compare with last_check timestamp in case if we have not run checker for a long time
if trigger.ttl and metric_state[
"timestamp"] + trigger.ttl < trigger.last_check[
"timestamp"]:
log.msg("Metric %s TTL expired for state %s" %
(t1.name, metric_state))
if trigger.ttl_state == state.DEL and metric_state.get(
"event_timestamp") is not None:
log.msg("Remove metric %s" % t1.name)
del check["metrics"][t1.name]
for tN, tName in time_series.other_targets_names.iteritems(
):
log.msg("Remove metric %s" % tName)
del check["metrics"][tName]
for pattern in trigger.struct.get("patterns"):
yield trigger.db.delPatternMetrics(pattern)
continue
time_series.update_state(
t1, check, state.to_metric_state(trigger.ttl_state),
None, trigger.last_check["timestamp"] - trigger.ttl)
yield event.compare_states(trigger,
metric_state,
t1.last_state,
metric_state["timestamp"],
metric=t1.name)
except StopIteration:
raise
except Exception:
log.err()
check["state"] = state.EXCEPTION
check["msg"] = "Trigger evaluation exception"
yield event.compare_states(trigger, check, trigger.last_check, now)
if trigger.update_score:
update_score(check)
yield trigger.db.setTriggerLastCheck(trigger.id, check)
def check(self, fromTime=None, now=None, cache_ttl=60):
now = now or int(time())
log.msg("Checking trigger %s" % self.id)
initialized = yield self.init(now, fromTime=fromTime)
if not initialized:
raise StopIteration
if fromTime is None:
fromTime = self.last_check.get("timestamp", now)
requestContext = datalib.createRequestContext(str(fromTime - (self.ttl or 600)), str(now))
check = {"metrics": self.last_check["metrics"].copy(), "state": state.OK, "timestamp": now}
try:
time_series = yield self.get_timeseries(requestContext)
for metric in requestContext['metrics']:
yield self.db.cleanupMetricValues(metric, now - config.METRICS_TTL,
cache_key=metric, cache_ttl=cache_ttl)
if len(time_series) == 0:
if self.ttl:
check["state"] = self.ttl_state
check["msg"] = "Trigger has no metrics"
yield self.compare_state(check, self.last_check, now)
else:
for t_series in time_series.values():
for tN in t_series:
check["metrics"][tN.name] = tN.last_state.copy()
for t1 in time_series[1]:
metric_state = check["metrics"][t1.name]
for value_timestamp in xrange(t1.start, now + t1.step, t1.step):
if value_timestamp <= t1.last_state["timestamp"]:
continue
expression_values = time_series.get_expression_values(t1, value_timestamp)
t1_value = expression_values["t1"]
if None in expression_values.values():
continue
expression_values.update({'warn_value': self.struct.get('warn_value'),
'error_value': self.struct.get('error_value'),
'PREV_STATE': metric_state['state']})
expression_state = expression.getExpression(self.struct.get('expression'),
**expression_values)
time_series.update_state(t1, check, expression_state, expression_values, value_timestamp)
yield self.compare_state(metric_state, t1.last_state,
value_timestamp, value=t1_value,
metric=t1.name)
# compare with last_check timestamp in case if we have not run checker for a long time
if self.ttl and metric_state["timestamp"] + self.ttl < self.last_check["timestamp"]:
log.msg("Metric %s TTL expired for state %s" % (t1.name, metric_state))
if self.ttl_state == state.DEL and metric_state.get("event_timestamp") is not None:
log.msg("Remove metric %s" % t1.name)
del check["metrics"][t1.name]
for tN, tName in time_series.other_targets_names.iteritems():
log.msg("Remove metric %s" % tName)
del check["metrics"][tName]
for pattern in self.struct.get("patterns"):
yield self.db.delPatternMetrics(pattern)
continue
time_series.update_state(t1, check, state.toMetricState(self.ttl_state), None,
self.last_check["timestamp"] - self.ttl)
yield self.compare_state(metric_state, t1.last_state, metric_state["timestamp"], metric=t1.name)
except StopIteration:
raise
except:
log.err()
check["state"] = state.EXCEPTION
check["msg"] = "Trigger evaluation exception"
yield self.compare_state(check, self.last_check, now)
yield self.db.setTriggerLastCheck(self.id, check)
