JuJu restore gets the new controller stuck in "juju restore is in progress" state
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Canonical Juju |
Triaged
|
Low
|
Unassigned |
Bug Description
I used the following command to restore the backup to another host
```juju restore-backup --id 20190220-
Which worked except for some permission errors that I troubleshot and discovered that the new host used agent machine-3 to spin up the juju-db and not machine-0 like the previous host and the new agent had no authority inside of mongo. I resolved that by duplicating the permissions from the original agent to the current one. Which resolved all the permission errors I received but started a new one when attempting to contact the controller and get its status.
```juju status --debug
09:19:58 INFO juju.cmd supercommand.go:56 running juju [2.3.7 gc go1.8]
09:19:58 DEBUG juju.cmd supercommand.go:57 args: []string{"juju", "status", "--debug"}
09:19:58 INFO juju.juju api.go:67 connecting to API addresses: [10.54.4.17:17070 10.54.4.2:17070 10.54.4.3:17070]
09:19:58 DEBUG juju.api apiclient.go:843 successfully dialed "wss://
09:19:58 INFO juju.api apiclient.go:597 connection established to "wss://
09:19:58 DEBUG juju.api monitor.go:35 RPC connection died
ERROR juju restore is in progress - API is disabled to prevent data loss
09:19:58 DEBUG cmd supercommand.go:459 error stack:
juju restore is in progress - API is disabled to prevent data loss
github.
github.
github.
I can find no solution for removing this server from its eternal restoring state, I remove the replset, i re-initialized the repleset, I attempted just about everything with the database I could think of. I cannot get the server out of this state and actually usable. This bug is rather frustrating when attempting to do restores. I duplicated the issue with a new controller.
Bootstrap new controller
place controller in HA - juju enable-ha
create backup
restore backup to new host
resolve permission issues
receive API disabled due to restore errors
Changed in juju: | |
status: | New → Triaged |
importance: | Undecided → High |
tags: | added: backup-restore |
tags: | added: restore-backup |
tags: | removed: restore-backup |
'juju restore-backup' is currently broken if the original backup was taken from an HA controller. This is due to a variety of reasons, mostly around the new controller isn't going to be HA, so a few of the restored bits bring up a configuration that says "you should be HA, wait for the other agents."
I have started working through a step-by-step to manually restore a controller, which we hope to incorporate into 'juju restore-backup'. I haven't finished polishing all the steps, but it might give you steps that you could follow.