Bug #20989
closedceph-mgr dashboard does not start on Linux container
0%
Description
- All MONs and MGRs run on Linux container (LXC, Ubuntu 16.04.3) on top of Proxmox VE 5.0 cluster.
- All OSDs run on 3 nodes of Proxmox VE 5.0 (based on Debian 9).
The Ceph cluster is running fine so far, except the Ceph MGR dashboard could not be started. The Ceph MGR dashboard always gets traceback error everytime I start the ceph-mgr service inside the container. Is there any limitation to run Ceph MON/MGR/MDS/RGW on Linux container?
Didn't have this issue if the Ceph MON and MGR are installed on physical machine though. I'll try to redeploy the MON and MGR to Linux KVM and see whether the Ceph MGR dashboard could be started or not.
2017-08-12 20:38:54.538945 7f3e02be5500 0 set uid:gid to 64045:64045 (ceph:ceph)
2017-08-12 20:38:54.538961 7f3e02be5500 0 ceph version 12.1.3 (c56d9c07b342c08419bbc18dcf2a4c5fae62b9cf) luminous (rc), process (unknown), pid 715
2017-08-12 20:38:54.538976 7f3e02be5500 0 pidfile_write: ignore empty --pid-file
2017-08-12 20:38:54.551844 7f3e02be5500 1 mgr send_beacon standby
2017-08-12 20:38:56.230155 7f3df9f19700 1 mgr handle_mgr_map Activating!
2017-08-12 20:38:56.230332 7f3df9f19700 1 mgr handle_mgr_map I am now activating
2017-08-12 20:38:56.311524 7f3df5f11700 1 mgr init Loading python module 'dashboard'
2017-08-12 20:38:56.397348 7f3df5f11700 1 mgr load Constructed class from module: dashboard
2017-08-12 20:38:56.397358 7f3df5f11700 1 mgr start Creating threads for 1 modules
2017-08-12 20:38:56.397403 7f3df5f11700 1 mgr send_beacon active
2017-08-12 20:38:56.398817 7f3de8cfe700 -1 mgr serve dashboard.serve:
2017-08-12 20:38:56.398821 7f3de8cfe700 -1 mgr serve Traceback (most recent call last):
File "/usr/lib/ceph/mgr/dashboard/module.py", line 831, in serve
'server.socket_port': int(server_port),
ValueError: invalid literal for int() with base 10: ''
2017-08-12 20:38:56.552933 7f3df6f13700 1 mgr send_beacon active
Updated by John Spray over 6 years ago
This looks a lot less likely to be anything to do with containers and a lot more likely to be a mistake setting socket_port during configuration. Did you set it to an empty string at some stage?
Updated by Charles Alva over 6 years ago
John Spray wrote:
This looks a lot less likely to be anything to do with containers and a lot more likely to be a mistake setting socket_port during configuration. Did you set it to an empty string at some stage?
Hi John,
I think did set it to blank before at 12.1.1 RC1 or so. I set it back to :: and 7000 for the server_addr and server_port respectively on 12.1.3 (last RC) and still could not start the dashboard.
# ceph config-key set mgr/dashboard/server_addr ::
# ceph config-key set mgr/dashboard/server_port 7000
# ceph config-key get mgr/dashboard/server_addr
obtained 'mgr/dashboard/server_addr'
::
# ceph config-key get mgr/dashboard/server_port
obtained 'mgr/dashboard/server_port'
7000
Updated by Charles Alva over 6 years ago
Whoa thanks for the hint. Just got it working after removing all the individual mgr/dashboard per server name settings and the cluster wide settings. I forgot I set per server name ceph mgr dashboard settings back then at 12.1.0 RC1 or so. Sorry for the false positive bug report. You may close this issue.
# ceph mgr module disable dashboard
# ceph config-key del mgr/dashboard/$name/server_addr
# ceph config-key del mgr/dashboard/$name/server_port
# ceph config-key del mgr/dashboard/server_addr
# ceph config-key del mgr/dashboard/server_port
# ceph mgr module enable dashboard
# systemctl restart ceph-mgr.target
Updated by John Spray over 6 years ago
- Status changed from New to Closed
Thanks for the update, closing