"juju deploy" destroys orphaned instances without warning
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
pyjuju |
Triaged
|
Low
|
Unassigned |
Bug Description
My instance 0 went away without leave. This was probably not juju's fault.
Having no controlling instance for my environment, I was forced to execute "juju bootstrap". This created a new machine 0. Unfortunately, it did not recognize the existing instances as being part of the environment. Those old instances were effectively orphaned at this point. I decided to tear down the orphaned instances, spin up new ones using "juju deploy ..." and migrate databases, and to do this one service at a time.
I destroyed the VMs for the first service, then executed "juju deploy ...". After a while, the new service was coming up, and I started receiving alerts that all of the others services had just gone down. Juju had terminated all of the old, orphaned instances, and done so without warning. This included a database.
Juju being able to clean up orphaned instances is a neat trick, but it really ought to be its own command, "juju cleanup" or something. If another juju command is unable to proceed because of orphaned instances in the way, it ought to inform the user of that problem, and advise use of that separate command.
Luckily, I make backups of complete database dumps every 5 minutes.
Changed in juju: | |
importance: | Undecided → Low |
status: | New → Triaged |
juju 0.6.0.1 from the PPA, on 12.10.