Closed tserong closed 1 year ago
Still inscrutable in the UI when testing #758, but the visuals are slightly different:
Here's what journalctl -u aquarium
had to say during my test, so the log is definitely helpful:
Dec 21 07:15:04 node1 uvicorn[1444]: INFO: 2021-12-21 07:15:04 -- create -- Bootstrap complete with success.
Dec 21 07:15:06 node1 uvicorn[1444]: ERROR: 2021-12-21 07:15:06 -- ceph -- error running command: rc = -2, reason = all mgr daemons do not support module 'bubbles', pass --force to force enablement
Dec 21 07:15:06 node1 uvicorn[1444]: ERROR: 2021-12-21 07:15:06 -- ceph -- unable to enable module bubbles: all mgr daemons do not support module 'bubbles', pass --force to force enablement
Dec 21 07:15:06 node1 uvicorn[1444]: ERROR: 2021-12-21 07:15:06 -- create -- Unable to start Bubbles.
Dec 21 07:15:06 node1 uvicorn[1444]: ERROR: 2021-12-21 07:15:06 -- create -- Create error: Failed configuring the deployment.
Dec 21 07:15:06 node1 uvicorn[1444]: INFO: 2021-12-21 07:15:06 -- create -- Waiting for task to finish.
Dec 21 07:15:06 node1 uvicorn[1444]: ERROR: 2021-12-21 07:15:06 -- mgr -- Error creating deployment: Failed configuring the deployment.
After you hit the "Install" button when creating a new cluster, Aquarium goes off and runs
cephadm bootstrap
. If this fails for some reason, all you see is a red box at the bottom of the screen which says "Failed to bootstrap the system", with no further information about what might be wrong:This error message disappears after a short amount of time, and your only option is to hit the "Install" button again and hope for a different outcome.
The aquarium log (
journalctl -u aquarium
) will give you something baffling, like:The actual cause of the problem can be found buried in /var/log/ceph/cephadm.log. In my case it was:
This was caused by me breaking my network configuration and having no default route, so of course
podman pull
can't do it's thing, but the problem is I had no idea why, until I went digging.IMO we need to:
1) Somehow propagate more useful failure information from
cephadm bootstrap
stdout/stderr/log/whatever to the UI 2) See if we can make /var/log/cephadm/cephadm.log any easier to read. AFAICT there's no indication in there of whichcephadm
command was invoked (bootstrap
,inventory
, ...) - instead you have to correlate with the aquarium log to see what was invoked, and when.(Item 2 is almost certainly something to address in cephadm itself and not specific to Aquarium)