Plugins » Katello

1. Problem

Two sources of redundant Candlepin traffic during host registration:

1. check_registration_services pre-flight check

Fires a Katello::Ping.ping call against Candlepin before every POST /rhsm/consumers, DELETE /rhsm/consumers/:id, and the equivalent host_subscriptions endpoints. The check is redundant because:
- rescue_from RestClient::Exception already handles Candlepin failures and returns the actual HTTP error code
- register_host already cleans up partial Foreman DB state when Candlepin fails mid-registration
- subscription-manager independently calls GET /rhsm/status ~12 times per registration
- A Candlepin in maintenance mode returns non-200 from the registration endpoint anyway

2. /rhsm/status proxied to Candlepin on every call

subscription-manager calls GET /rhsm/status approximately 12 times per registration, all proxied to Candlepin /candlepin/status. Under 100 concurrent registrations this generates ~1200 Candlepin status calls competing with actual registration traffic.

1. Fix

1. Remove check_registration_services entirely from both candlepin_proxies_controller and host_subscriptions_controller. The host_subscriptions_controller gains an explicit rescue_from RestClient::Exception that preserves 4xx semantics and returns 503 only for genuine backend unavailability.

2. Cache the /rhsm/status Candlepin response in server_status for 30 seconds using Rails.cache.fetch. The first call per window hits Candlepin; subsequent calls are served from cache. Under 100 concurrent registrations this reduces ~1200 Candlepin status calls to approximately 1 per 30-second window.