ben/fast_ilo_exporter
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

improve error handling

Browse Source
This commit is contained in:
Benjamin Wiegand 2024-05-20 00:52:16 -07:00 committed by Benjamin Wiegand
parent 01c44aa2fe
commit ff206c24cb
1 changed files with 35 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

71
main.py
View File

@ -17,7 +17,6 @@ from targets.memory import *
import targets.power import targets.power
import argparse import argparse
import traceback
NAMESPACE = 'ilo' NAMESPACE = 'ilo'
@ -41,7 +40,7 @@ if args.quiet and args.verbose:
print('stop it. (--quiet and --verbose do not mix)') print('stop it. (--quiet and --verbose do not mix)')
exit(1) exit(1)
SCAN_FAIL_COUNTER = Counter('exporter', 'Number of times scanning the iLO for SNMP variables has failed.', namespace=NAMESPACE, subsystem='snmp_scan_failures') SCAN_FAIL_COUNTER = Counter('scrape_failures', 'Number of times scraping the iLO for SNMP variables has failed.', namespace=NAMESPACE, subsystem='exporter')
def noisy(*a, **kwa): def noisy(*a, **kwa):
@ -76,49 +75,49 @@ class BulkCollector(Collector):
def collect(self): def collect(self):
cache = {} cache = {}
if self._scan_on_collect: try:
try: if self._scan_on_collect:
self.scan() self.scan()
except Exception as e:
traceback.print_exception(e)
print('Failed to scan SNMP, aborting collection')
SCAN_FAIL_COUNTER.inc()
return
for documentation, bulk_values, bulk_labels in self._metrics_groups: for documentation, bulk_values, bulk_labels in self._metrics_groups:
metric_name = self._name_template % bulk_values.name metric_name = self._name_template % bulk_values.name
verbose('collecting', metric_name) verbose('collecting', metric_name)
label_names = ['id'] label_names = ['id']
label_maps = [] label_maps = []
for label in bulk_labels: for label in bulk_labels:
# the labels are cached since they may be reused # the labels are cached since they may be reused
if label.name not in cache: if label.name not in cache:
cache[label.name] = label.get_values(self._snmp_config, self._ids) cache[label.name] = label.get_values(self._snmp_config, self._ids)
label_names.append(label.name) label_names.append(label.name)
label_maps.append(cache[label.name]) label_maps.append(cache[label.name])
metric = GaugeMetricFamily( metric = GaugeMetricFamily(
metric_name, metric_name,
documentation, documentation,
labels=label_names labels=label_names
) )
# values are not reused # values are not reused
value_map = bulk_values.get_values(self._snmp_config, self._ids) value_map = bulk_values.get_values(self._snmp_config, self._ids)
# do some fuckery (bad design, I know.) # do some fuckery (bad design, I know.)
for i in self._ids: for i in self._ids:
labels = [str(i)] # id is first labels = [str(i)] # id is first
for label_map in label_maps: for label_map in label_maps:
label_value = label_map[i] label_value = label_map[i]
labels.append(str(label_value)) labels.append(str(label_value))
value = value_map[i] value = value_map[i]
metric.add_metric(labels, value) metric.add_metric(labels, value)
yield metric yield metric
except Exception as e:
print('Failed to scan SNMP, aborting collection')
SCAN_FAIL_COUNTER.inc()
raise e
def get_power_draw() -> float: def get_power_draw() -> float: