[pve-devel] [PATCH container v2] Fix #576: Cancelling move disk does not leave useless files anymore

[Thomas Lamprecht]

On 6/12/19 11:10 AM, Dominic Jäger wrote:
> I'll make more detailed commit messages from now on.
> 
> On 6/12/19 7:59 AM, Thomas Lamprecht wrote:
>> On 6/6/19 11:52 AM, Dominic Jäger wrote:
>>> When cancelling the move disk operation for containers the partly
>>> finished destination and thus useless files now get removed.
>> I really do not understand the commit message, could you please
>> describe here how/why that happens now?
> 
> When Move Disk is called for a container rsync starts copying it to a new destination. This initial rsync process gets killed when the Stop button gets pressed. At this moment the destination file is not fully copied and useless as a consequence. Our code already tries to remove it. However, rsync has forked and those forks are still accessing the destination file for some time. Thus, the attempt to remove it fails.
> 
> With the patch we wait for other processes (which are only rsync's children normally) to release the destination files. This took less than a second when I tried it. Afterwards, the existing remove procedure is carried out.
> 

Now that sound like a much better commit message ;-)
Maybe throw in that it runs in a mount namespace and is, as Wolfgang
correctly said, protected by our config lock, so there's no possibility
that we interrupted another, concurrent, operation of the PVE stack here.