Page MenuHomePhabricator

Prometheus doing requests via proxy that gets 400
Open, Needs TriagePublic

Description

While deploying a squid patch I was monitoring the squid logs and noticed that there were a bunch of requests getting 400s from the prometheus hosts, see: https://logstash.wikimedia.org/goto/3c89a5c370e6a5531b61c955bbd3cf44

Extract from the logs:

"http.request.method": "GET",
"url.full": "/",
"url.path": "-",
"url.domain": "-",
"http.response.status_code": 400,
"http.response.bytes": 3591,
"http.version": "1.1",
"http.request.referrer": "-",
"user_agent.original": "-",

They all seem to come from the prometheys hosts, missing a UA and the url domain.

Event Timeline

This has been done on purpose (see modules/profile/manifests/installserver/proxy.pp:65). I believe another approach, to check for a 200 OK, could have been to allow Prometheus hosts, via an ACL, to query the :8080/squid-internal-mgr/info endpoint.