Description of problem: The lvm2-dbusd API returns Job objects for actions being performed. The Job objects provide results, progress and other features as described at https://github.com/tasleson/lvm-dubstep/blob/master/api.md#interface-comredhatlvmdbus1job However, there's no way to cancel a Job. For example a job for 'pvmove' can take a very log time while the underlying operation can be safely canceled. It would be nice if the com.redhat.lvmdbus1.Job interface provided a Cancel() method or there was a new com.redhat.lvmdbus1.Job.Cancellable interface providing such method at least for specific types of actions. (If the Cancel() method is added to the com.redhat.lvmdbus1.Job interface, it may just return error if the action it runs doesn't support cancellation)
The only thing we can interrupt is pvmove(s). However, lvm only gives us the ability to abort *all* outstanding moves, not specific ones. I will follow-up with the lvm team on if we could provide a mechanism to abort specific move operations. Otherwise I think we are only left with something like Manager.PvMoveAbort().
The command line is supposed to work in exactly the same way as other options - if you give additional arguments it's supposed to use them to identify the pvmove(s) the action applies to. Is this facility missing from the API?
(In reply to Alasdair Kergon from comment #2) > The command line is supposed to work in exactly the same way as other > options - if you give additional arguments it's supposed to use them to > identify the pvmove(s) the action applies to. OK I incorrectly concluded from the man page: "--abort Abort any moves in progress. If the --atomic option was used to start the pvmove, all logical volumes will remain on the source physical vol‐ ume. Otherwise, those segments that have completed the move will stay on the destination physical volume, while those that have not will remain on the source physical volume." that `pvmove --abort` was the only option (which does indeed work). In empirical testing it appears that you can select a specific move to abort. > Is this facility missing from > the API? Yes, thus the reason for this specific BZ
> OK I incorrectly concluded from the man page:... Same here. Looks like the man page needs to be more explicit.