descIntegrationDevice was registered with namespace prefix
"unpoller_device_", producing "unpoller_device_uptime_seconds" with
labels {device_id} and a different help string than the existing
descDevice() metric of the same FQDN (labels {type, site_name, name,
source, tag}, help "Device Uptime"). Prometheus MustRegister panics on
inconsistent descriptors for the same fully-qualified name, causing
v3.0.0 to crashloop on startup whenever the Prometheus output was
enabled.
Move the Integration/v1 device metrics under a dedicated
"integration_device_" name prefix, matching the convention used by the
other Integration/v1 collectors added in the same release (e.g.
wifi_broadcast_*, acl_rule_*, mclag_domain_*, pending_device_*), where
the bare namespace prefix is passed in and the type prefix is baked
into each metric name string.
Affected metric renames:
unpoller_device_uptime_seconds -> unpoller_integration_device_uptime_seconds
unpoller_device_cpu_utilization_pct -> unpoller_integration_device_cpu_utilization_pct
unpoller_device_memory_utilization_pct -> unpoller_integration_device_memory_utilization_pct
unpoller_device_load_average_{1,5,15}min -> unpoller_integration_device_load_average_{1,5,15}min
unpoller_device_radio_tx_retries_pct -> unpoller_integration_device_radio_tx_retries_pct
unpoller_device_uplink_{rx,tx}_rate_bps -> unpoller_integration_device_uplink_{rx,tx}_rate_bps
Fixes#1002Fixes#1004
c48b9917b0