hw-health charm can hold onto bad sensor data

[Adam Dyess]

I found a case where sensor data on a disk was reporting fault:

$ sudo ipmi-sensors | egrep -i 'DISK45|DISK44'
52 | SSD Disk45 Temp | Temperature | 0.00 | C | 'At or Below (<=) Lower Non-Recoverable Threshold'
53 | SSD Disk44 Temp | Temperature | 78.00 | C | 'OK'
176 | DISK44 | Drive Slot | N/A | N/A | 'Drive Presence'
177 | DISK45 | Drive Slot | N/A | N/A | 'Drive Presence'

Notice the Disk45 temperature at an icy 0.00 C. the bmc reported a more moderate 55 C for both drives actually, and it turned out the ipmi sensor data is cached at

/root/.freeipmi/sdr-cache/sdr-cache-$(hostname)

The charm should evaluate if this sensor data is changing over time similar to the way it detects stale data in /var/lib/nagios/ipmi_sensors.out and clear the cache if necessary.

Secondly, the charm should provide an action to clear this cache on demand

Workaround:
`sudo ipmi-sensors -f`

[Andrea Ieri (aieri)]

This is odd, the check_ipmi_sensor script we rely on is already invoking ipmi-sensors with the --sdr-cache-recreate option (see https://github.com/thomas-krenn/check_ipmi_sensor_v3/blob/master/check_ipmi_sensor#L753), so a stale cache should automatically be detected and handled.

Could this be a bug in freeipmi?
Is the cache recreated if `sudo ipmi-sensors --sdr-cache-recreate` is run manually?

[Launchpad Janitor (janitor)]

[Expired for hw-health-charm because there has been no activity for 60 days.]