matthew/sonic-buildimage
Archived
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[dell]: System Health: Fix ASIC key issue in Dell platform (#6556)

Browse Source 
ASIC key used in system health daemon is not present in Dell platforms.

Fixes #6343

Got the thermal sensor list using 2.0 API and retrieved the ASIC keys.
This commit is contained in:
Aravind Mani 2021-04-06 06:30:38 +05:30 committed by GitHub
parent 7aa03bead9
commit 6d83a424b5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 29 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
src/system-health/health_checker/hardware_checker.py
View File

@ -8,6 +8,7 @@ class HardwareChecker(HealthChecker):
"""
Check system hardware status. For now, it checks ASIC, PSU and fan status.
"""
ASIC_TEMPERATURE_KEY = 'TEMPERATURE_INFO|ASIC'
FAN_TABLE_NAME = 'FAN_INFO'
PSU_TABLE_NAME = 'PSU_INFO'
@ -35,27 +36,34 @@ class HardwareChecker(HealthChecker):
if config.ignore_devices and 'asic' in config.ignore_devices:
return
temperature = self._db.get(self._db.STATE_DB, HardwareChecker.ASIC_TEMPERATURE_KEY, 'temperature')
temperature_threshold = self._db.get(self._db.STATE_DB, HardwareChecker.ASIC_TEMPERATURE_KEY, 'high_threshold')
if not temperature:
self.set_object_not_ok('ASIC', 'ASIC', 'Failed to get ASIC temperature')
elif not temperature_threshold:
self.set_object_not_ok('ASIC', 'ASIC', 'Failed to get ASIC temperature threshold')
else:
try:
temperature = float(temperature)
temperature_threshold = float(temperature_threshold)
if temperature > temperature_threshold:
self.set_object_not_ok('ASIC', 'ASIC',
'ASIC temperature is too hot, temperature={}, threshold={}'.format(
temperature,
temperature_threshold))
else:
self.set_object_ok('ASIC', 'ASIC')
except ValueError as e:
self.set_object_not_ok('ASIC', 'ASIC',
'Invalid ASIC temperature data, temperature={}, threshold={}'.format(temperature,
temperature_threshold))
ASIC_TEMPERATURE_KEY_LIST = self._db.keys(self._db.STATE_DB,
HardwareChecker.ASIC_TEMPERATURE_KEY + '*')
for asic_key in ASIC_TEMPERATURE_KEY_LIST:
temperature = self._db.get(self._db.STATE_DB, asic_key,
'temperature')
temperature_threshold = self._db.get(self._db.STATE_DB, asic_key,
'high_threshold')
asic_name = asic_key.split('|')[1]
if not temperature:
self.set_object_not_ok('ASIC', asic_name,
'Failed to get {} temperature'.format(asic_name))
elif not temperature_threshold:
self.set_object_not_ok('ASIC', asic_name,
'Failed to get {} temperature threshold'.format(asic_name))
else:
try:
temperature = float(temperature)
temperature_threshold = float(temperature_threshold)
if temperature > temperature_threshold:
self.set_object_not_ok('ASIC', asic_name,
'{} temperature is too hot, temperature={}, threshold={}'.format(
asic_name, temperature, temperature_threshold))
else:
self.set_object_ok('ASIC', asic_name)
except ValueError as e:
self.set_object_not_ok('ASIC', asic_name,
'Invalid {} temperature data, temperature={}, threshold={}'.format(
asic_name, temperature, temperature_threshold))
def _check_fan_status(self, config):
"""