Override for grafana_server_addr (fact derived)
See original GitHub issueIs your feature request related to a problem? Please describe. Would like to have the ability to override fact-derived value for grafana_server_addr. Current playbook fact is derived from interface falling into the range for public_network.
Describe the solution you’d like My deployment contains public and separate cluster networks into non-routable, isolated subnets. Each monitor, OSD, etc. are dual homed to both the public, optionally cluster networks as well as connected to a management network (this one is routed, default gw defined). When connecting ceph-ansible deployed Grafana and connecting to deployed ceph-mgr, dashboard is unable to display Grafana dashboards since iframe is redirecting directly to grafana_server_addr:3000 – in my case is not routable from client viewing dashboard.
Describe alternatives you’ve considered Manually overriding via:
ceph dashboard set-grafana-api-url http://(custom IP):3000
does the trick. However has to be done every time Ceph-ansible overrides the value.
Issue Analytics
- State:
- Created 4 years ago
- Reactions:1
- Comments:9 (3 by maintainers)
Top GitHub Comments
This would be simpler if the ceph-mgr dashboard proxied grafana requests vs. referencing them. In my particular setup public is an isolated /protected network anyway, so no issues with unauthorized (ie public) grafana access. ideally, grafana should be authenticated, and proxied. However I gather that’s not the scope of ceph-ansible.
@dsavineau a simple override as suggested above (
grafana_server_addr
)as part of the playbooks would be more than sufficient.Seconded.
Since we are using Ceph purely for VMware shared storage, we have segregated our Ceph networks completely; they aren’t even privately route-able. We have an additional management interface which is firewalled off but gives authorized users access to the Ceph management GUI and the machines. This setup works great except for Grafana, which binds to the unroutable Ceph subnet.
Passing grafana_server_addr as an Ansible
-x
(Edit:-e
I believe. It’s been a while) parameter fixes this, but is annoying.