fleetctl (0.11.5) destroy is not backward compatible

[btalbot]

Older versions of fleetctl seems to ignore the request to destroy units which do not actually exist. This was a handy command to use for startup / provisioning / startup scrips to stop and destroy a unit that may or may not actually be loaded.

core@ip-172-11-22-333 ~ $ fleetctl --version
fleetctl version 0.10.2
core@ip-172-11-22-333 ~ $ fleetctl destroy foo
core@ip-172-11-22-333 ~ $ echo $?
0

However, fleetctl 0.11.5 causes destroy show an unhelpful error message and exit with code 1 if the unit is not loaded

core@core-01 ~ $ fleetctl --version
fleetctl version 0.11.5
core@core-01 ~ $ fleetctl destroy fubar
Error destroying units: googleapi: Error 404: unit does not exist
core@core-01 ~ $ echo $?
1

This has now hit the beta channel and broke some of our development and test deployment scripts.

Is the new behavior a regression, bug, or a feature?

[mischief]

this was introduced by 9530eed

[btalbot]

Note that other fleetctl commands for units that do not exist do not generate the error message or exit code:

core@core-01 ~ $ fleetctl --version
fleetctl version 0.11.5
core@core-01 ~ $ fleetctl stop fubar; echo $?
0
core@core-01 ~ $ fleetctl unload fubar; echo $?
0

[megahall]

Would it be possible to backport the fix in #1417 into Fleet v0.11.x or ship a CoreOS which includes Fleet 0.12.x? Because no current version of CoreOS includes Fleet 0.12.x as of yet.

[dongsupark]

@megahall 👍
However this issue is already closed. How about creating a new issue on https://github.com/coreos/bugs/issues, so that other maintainers could easily track it?

[megahall]

Downstream report filed as coreos/bugs#1448 .