[checkmk-commits] agent_azure: improve handling of metrics

Module: check_mk Branch: master Commit: 45d53c03e55142246ab84786119a9684c9ff64d3 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=45d53c03e55142… Author: Moritz Kiemer &lt;mo(a)mathias-kettner.de&gt; Date: Fri Dec 21 11:50:07 2018 +0100 agent_azure: improve handling of metrics * only one aggregation per metric * output both expected and true interval, if we can determine it Change-Id: I4bf7880d7267d237e07e0d3b96e2d635e31431f5 --- agents/special/agent_azure | 132 +++++++++++++++++++++++++-------------------- 1 file changed, 74 insertions(+), 58 deletions(-) diff --git a/agents/special/agent_azure b/agents/special/agent_azure index f82183d..df915c4 100755 --- a/agents/special/agent_azure +++ b/agents/special/agent_azure @@ -365,45 +365,53 @@ class IssueCollecter(object): return len(self._list) -class AzureMetric(object): # pylint: disable=too-few-public-methods +class AzureMetricParseError(ValueError): + pass - HEADER = ("name", "aggregation", "value", "unit", "timestamp", "timegrain", "filters") - AGGREGATIONS = ("total", "average", "maximum", "minimum", "last", "count") - def __init__(self, metric, aggregations, time_grain, filter_, rid, err): +class AzureMetric(object): # pylint: disable=too-many-instance-attributes + + HEADER = ("name", "aggregation", "value", "unit", "timestamp", "filter", "interval_id", + "interval") + + def __init__(self, metric, aggregation, interval_id, filter_): super(AzureMetric, self).__init__() - self.rows = [] self.name = metric.name.value + self.aggregation = aggregation + self.value = None + self.unit = metric.unit.name + self.timestamp = None + self.filter = filter_ + self.interval_id = interval_id + self.interval = None - # Assuming this agent is executed to gather one value per metric count = len(metric.timeseries) - if count <> 1: - msg = "timeseries has %d elements for metric %s" % (count, self.name) - err.add("info", rid, msg) - LOG.warning(msg + " for resource %s" % rid) + if count == 0: + msg = "no timeseries found for metric %r" % self.name + raise AzureMetricParseError('info', msg) for measurement in reversed(metric.timeseries): if not measurement.data: continue - for aggr in aggregations.split(','): - for data in reversed(measurement.data): - if aggregations.startswith(aggr): - LOG.debug("data: %s", data) - if hasattr(data, aggr): - value = getattr(data, aggr) - # the field *may* be present, even if we did not ask for it. - if value is not None: - timestamp = calendar.timegm(data.time_stamp.utctimetuple()) - self.rows.append((self.name, aggr, value, metric.unit.name, timestamp, - time_grain, filter_)) - break - else: # no break ocurred, we did not find anything - logging.debug("not found: %s %s", self.name, aggr) - if not self.rows: - LOG.info("Found no aggregation for metric %r", self.name) - return + try: + self.interval = (measurement.data[-1].time_stamp - measurement.data[-2].time_stamp) + except (IndexError, TypeError): + pass + + for data in reversed(measurement.data): + LOG.debug("data: %s", data) + self.value = getattr(data, aggregation, None) + if self.value is not None: + self.timestamp = calendar.timegm(data.time_stamp.utctimetuple()) + return + + raise AzureMetricParseError('warning', "not found: %s %s", self.name, aggregation) + + @property + def tuple(self): + return tuple(getattr(self, field) for field in AzureMetric.HEADER) class AzureResource(object): @@ -424,9 +432,9 @@ class AzureResource(object): def dumpinfo(self): lines = [("Resource",), (json.dumps(self.info),)] - m_lines = sum((m.rows for m in self.metrics), []) - if m_lines: - lines += [("metrics following", len(m_lines)), self.metrics[0].HEADER] + m_lines + if self.metrics: + lines += [("metrics following", len(self.metrics)), AzureMetric.HEADER] + lines += [m.tuple for m in self.metrics] return lines @@ -508,44 +516,52 @@ class AzureClient(object): # pylint: disable=too-many-instance-attributes self.resources_cache = map(AzureResource, raw_resources) return self.resources_cache - def _fetch_specific_metrics(self, resource, metricnames, time_grain, aggregations, filter_, - err): + def _metric_api_call(self, rid, timespan, interval, metric, aggregation, filter_): + LOG.debug( + "metrics.list(%r, timespan=%r, interval=%r, metric=%r, aggregation=%r," + " filter=%r, raw=True)", rid, timespan, interval, metric, aggregation, filter_) + raw = self._monitor_client.metrics.list( + rid, + timespan=timespan, + interval=interval, + metric=metric, + aggregation=aggregation, + filter=filter_, + raw=True, + ) + # raw.output is what we'd gotten had we set raw=False. + # It is a paged object, make sure to actually retrieve + # all pages (as this may raise exceptions) + raw_metrics = list(raw.output.value) + return raw.response, raw_metrics + + def _fetch_specific_metrics(self, resource, metricnames, time_grain, aggregation, filter_, err): if metricnames.count(',') >= 20: raise ValueError("Azure API won't have requests with more than 20 metrics!") - metrics = [] - rid = resource.info["id"] timespan = self.timespans[time_grain] - LOG.debug( - "metrics.list(%r, timespan=%r, interval=%r, metricnames=%r, aggregation=%r," - " filter=%r, raw=True)", rid, timespan, time_grain, metricnames, aggregations, filter_) try: - # azure-api-call - raw = self._monitor_client.metrics.list( - rid, - timespan=timespan, - interval=time_grain, - metric=metricnames, - aggregation=aggregations, - filter=filter_, - raw=True, - ) - LOG.debug("raw.response: %s", raw.response) - self.remaining_reads = raw.response.headers[ - 'x-ms-ratelimit-remaining-subscription-reads'] - # raw.output is what we'd gotten had we set raw=False - LOG.debug("raw.ouput: %s", raw.output) - metrics += [ - AzureMetric(metric, aggregations, time_grain, filter_, rid, err) - for metric in raw.output.value - ] + response, raw_metrics = self._metric_api_call(rid, timespan, time_grain, metricnames, + aggregation, filter_) except ErrorResponseException as exc: if self.args.debug: raise exc err.add("exception", rid, exc.message) LOG.exception(exc) + return [] + + LOG.debug("response: %s", response) + self.remaining_reads = response.headers['x-ms-ratelimit-remaining-subscription-reads'] + + metrics = [] + for raw_metric in raw_metrics: + try: + metrics.append(AzureMetric(raw_metric, aggregation, time_grain, filter_)) + except AzureMetricParseError as exc: + err.add(exc[0], rid, exc[1]) + LOG.warning(exc.args[1]) return metrics @@ -556,8 +572,8 @@ class AzureClient(object): # pylint: disable=too-many-instance-attributes metric_params = METRICS_SELECTED.get(resource.info["type"], []) metrics = [] - for metricnames, time_grain, aggregations, filter_ in metric_params: - metrics += self._fetch_specific_metrics(resource, metricnames, time_grain, aggregations, + for metricnames, time_grain, aggregation, filter_ in metric_params: + metrics += self._fetch_specific_metrics(resource, metricnames, time_grain, aggregation, filter_, err) return metrics