Expose monitoring threshold/severity for "slow requests"

[Drew Freiberger]

The following alert from the nagios check_ceph is classified as a Warning, but should be Critical as it will typically require operator intervention to resolve.

396 slow requests are blocked > 32 sec (REQUEST_SLOW)

Can you please expose a configuration to allow for setting the alert severity of this to Critical?

Current workaround is to perform the following config change:

juju config ceph-mon nagios_additional_checks_critical=true nagios_additional_checks='{"ceph_slow_requests": "slow requests"}'

[Drew Freiberger (afreiberger)]

Adding juju-lint project, as it should check for existence of ceph-mon nagios_additional_checks_critical=true and nagios_additional_checks='{"ceph_slow_requests": "slow requests"}' until this bug is addressed within ceph-mon charm.

[Peter Jose De Sousa (pjds)]

The fix provided via juju lint appears to not be working as expected. The values will be parsed down to the nagios python check file as:

`--additional-checks "'slow' 'requests'"` which fails consitently, the correct value would be --additional-checks 'slow requests'

[Peter Sabaini (peter-sabaini)]

Going to mark this as wontfix as we're transitioning away from LMA. The new COS based metrics and alerting does support checking for blocked pgs https://discourse.ubuntu.com/t/metrics-and-alerts/45526