prometheus metrics output

[anarcat]

Hi!

We're (possibly) transitioning away from Icinga to Prometheus for our monitoring down here and it would be quite nice to have the equivalent functionality to Icinga, but as an OpenMetrics endpoint.

I am not exactly sure what the metrics would be like. It seems to me there could be different metrics for kernel, ucode, and services, possibly with a separation between user and system services. So something like this, maybe:

# HELP needrestart_timestamp information about the running version and when it was last updated
# TYPE needrestart_timestamp gauge
needrestart_timestamp{version=3.6} 1700675409
# HELP needrestart_kernel_info information about the kernel
# TYPE needrestart_kernel_info info
needrestart_kernel_info{running=6.5.0-1-amd64,expected=6.5.0-1-amd64,status="current"} 1
# HELP needrestart_ucode_info information about the CPU microcode
# TYPE needrestart_ucode_info info
needrestart_ucode_info{running=0x042c,expected=0x042c,status="current"} 1
# HELP needrestart_services_count number of services requiring a restart
# TYPE needrestart_services_count gauge
needrestart_services_count = 3

It would probably need gauges for containers and sessions too...

Would people here be open to this idea?

[anarcat]

Note that there's some overlap between this and the node exporter's support for such thing. This was requested in prometheus/node_exporter#625 but actually implemented in the "collectors" project. It only tracks the reboot-required file, however...

[liske]

I'm open for changes required for a metrics endpoint. :+1:

[tewfik-ghariani]

I just came across this project that might be a good solution: https://git.fsmpi.rwth-aachen.de/thomas/needrestart2prom/-/tree/main