Fix for prometheus scraping of cAdvisor metrics (#957)

* exposing cadvisor metrics on port 48080 for prometheus scraping

* update comment

---------

Co-authored-by: Jacob Pleiness <jdpleiness@users.noreply.github.com>

Author: loujar

docker-compose/docker-compose.yaml

@@ -486,7 +486,7 @@ services:
   # Description: Publishes Prometheus metrics about Docker containers.
   #
   # Disk: none
-  # Ports exposed to other Sourcegraph services: 8080/TCP
+  # Ports exposed to other Sourcegraph services: 48080/TCP
   # Ports exposed to the public internet: none
   #
   cadvisor:
@@ -513,8 +513,8 @@ services:
     networks:
       - sourcegraph
     restart: always
-    command:
-      - --port=8080
+    expose:
+      - 48080
 
   # Description: Publishes Prometheus metrics about the machine's hardware / operating system.
   #

prometheus/prometheus_targets.yml

@@ -3,7 +3,7 @@
     nodename: "sourcegraph-docker-compose-host"
     job: node
   targets:
-    - cadvisor:8080
+    - cadvisor:48080
     - sourcegraph-frontend-internal:6060
 - labels:
     nodename: "sourcegraph-docker-compose-host"