Fix race create sessione su CPX: ready solo dopo MediaMTX.

I nodi cloud restano in provisioning finché :9997 risponde; retry su create_path e 503 retryable se l’ingest è ancora irraggiungibile.

Co-authored-by: Cursor <cursoragent@cursor.com>

backend/app/services/mediamtx/client.rb

@@ -4,6 +4,9 @@ module Mediamtx
   class Client
     class Error < StandardError; end
 
+    CREATE_PATH_RETRIES = -> { ENV.fetch("MEDIAMTX_CREATE_RETRIES", "5").to_i }
+    CREATE_PATH_RETRY_BASE_SECS = -> { ENV.fetch("MEDIAMTX_CREATE_RETRY_BASE_SECS", "0.4").to_f }
+
     def self.for_session(session)
       new(base_url: session.mediamtx_api_base_url)
     end
@@ -19,6 +22,19 @@ module Mediamtx
 
     attr_reader :base_url
 
+    # Health probe for CPX readiness (GET /v3/paths/list).
+    def reachable?(timeout: 2)
+      conn = Faraday.new(url: @base_url) do |f|
+        f.adapter Faraday.default_adapter
+        f.options.open_timeout = timeout
+        f.options.timeout = timeout
+      end
+      response = conn.get("/v3/paths/list")
+      response.success?
+    rescue Faraday::Error
+      false
+    end
+
     def create_path(session)
       path = session.mediamtx_path_name
       # record: false finché non c'è publisher — con alwaysAvailable MediaMTX registrerebbe
@@ -30,7 +46,9 @@ module Mediamtx
       body[:alwaysAvailable] = true
       body[:alwaysAvailableFile] = slate_file_path(session)
       # YouTube: telefono → MediaMTX; relay copy verso RTMPS in sidekiq.
-      response = @conn.post("/v3/config/paths/add/#{CGI.escape(path)}", body)
+      response = with_connection_retries("create_path #{path}") do
+        @conn.post("/v3/config/paths/add/#{CGI.escape(path)}", body)
+      end
       unless response.success?
         err = response.body.is_a?(Hash) ? response.body["error"] : response.body
         raise Error, "MediaMTX path create failed: #{response.status} #{err}"
@@ -161,6 +179,23 @@ module Mediamtx
 
     private
 
+    def with_connection_retries(label)
+      attempts = [CREATE_PATH_RETRIES.call, 1].max
+      base = CREATE_PATH_RETRY_BASE_SECS.call
+      try = 0
+      begin
+        try += 1
+        yield
+      rescue Faraday::ConnectionFailed, Faraday::TimeoutError => e
+        raise if try >= attempts
+
+        sleep_secs = base * (2**(try - 1))
+        Rails.logger.warn("[Mediamtx::Client] #{label} retry #{try}/#{attempts} after #{e.class}: #{e.message} (sleep #{sleep_secs}s)")
+        sleep(sleep_secs)
+        retry
+      end
+    end
+
     def recording_body(session, enabled:)
       ent = session.match.team.entitlements
       can_record = ent.recording_enabled_for_mediamtx?