Ceph » mgr » Dashboard

Description of problem
When the RGW is behind a proxy and not reachable from outside the dashboard rgw discovery logic returns the wrong port and ssl settings. In our case the MGR runs on a bare metal machine and the rgw in a container on k8s non ssl port 8080. The rgw is reachable via k8s ingress with ssl on port 443. But since the mgr parses the rgw frontend config it tries to connect on port 8080 without ssl instead of 443 ssl.¶

But the same also applies if the rgw listens non ssl localhost on a normal machine, behind a reverse proxy which does the ssl termination.

Environment

• ceph version string: ceph version 17.2.7 (b12291d110049b2f35e32e0de30d70e9a4c060d2) quincy (stable)
• Platform (OS/distro/release): ceph docker image and Oracle Linux 8.8
• Cluster details (nodes, monitors, OSDs):
  RGW runs in k8s and the ceph cluster(mon,osd) outside on bare metal

• Did it happen on a stable environment or after a migration/upgrade?:
  it's a bug since the auto discovery of the rgw was introduced, in the past it worked fine with the: ceph dashboard set-rgw-api-host/set-rgw-api-scheme/set-rgw-api-port

Additional info¶

I think the only way to solve every case of the auto discovery problem is to have a config to override every option.