Add Prometheus monitoring endpoint #28
Conversation
|
Sorry for the delay; I've now added a middleware to have include metrics on the syscalls made and a small histogram. Note this is useful even without the middleware, just being able to monitor the process is running via metrics and the Go level metrics can be useful. |
Currently this has no specific metrics for seccompagent, aside from Go built-in build info.
This adds an optional middleware to get Prometheus metrics for request latencies as a histogram.
|
@alban can you please take a look again? |
| }).Error("Error in decoding syscall") | ||
| } | ||
|
|
||
| r := h(fd, req) |
There was a problem hiding this comment.
I think this is not measuring the duration of the syscall by the target process, but how long the seccomp agent takes to emulate the syscall (if such a handler is used for emulating the syscall).
I am curious about your use case and what syscall you emulate.
There was a problem hiding this comment.
The main one is allowing mount, we have a few more changes (I will open PRs for those soon) to allow network filesystems and that's definitely a place where these metrics are useful.
Add Prometheus monitoring endpoint
Adds an optional HTTP server, with a metrics endpoint available on
/metrics. This allows for Prometheus compatible metrics to be collected; currently it adds no metrics specific to seccompagent, but those could be added later.This adds a dependency on github.com/prometheus/client_golang, I've updated vendoring as a result (in a different commit, ideally don't squash merge this to keep the history clearer).
How to use
Run seccompagent with the monitor listen option set to a port.
Testing done