[CRIU] fsnotify: Can't open file handle for 0x00800001:0x0000000000027570: Stale file handle

[Pavel Emelyanov]

On 04/23/2014 05:56 PM, linuxchyu at sina.com wrote:
> 
> Hello， I caught an error when i do the migration operation of LXC.

Are you doing it by hands or using the https://github.com/xemul/p.haul?

> I copy all the disk files and dumped images of a CT to a dst host, when i do the
> restore operation in the dst host, some errors happened (there is no such error in
> the src host to do the same restore operation), and error logs are :
> (00.797923) 1: Restoring size 0x10000 for 0x7b6f
> (00.797950) 1: Wait fdinfo pid=1 fd=4
> (00.797965) 1: Send fd 5 to /crtools-fd-1-4
> (00.798030) 1: Create fd for 3
> (00.798057) 1: Restoring fd 4 (state -> create)
> (00.798070) 1: Creating pipe pipe_id=0x7b6f id=0x3
> (00.798080) 1: Waiting fd for 4
> (00.798131) 1: Create fd for 4
> (00.798155) 1: Restoring fd 5 (state -> create)
> (00.798182) 1: fsnotify: Restore 1 wd for 0x00000004
> (00.798220) 1: fsnotify: Opening fhandle 800001:3ef4955800027570...
> (00.798278) 1: Error (fsnotify.c:124): fsnotify: Can't open file handle for 0x00800001:0x0000000000027570: Stale file handle
> (00.845082) Error (cr-restore.c:1036): 4087 exited, status=255
> (00.845134) Error (cr-restore.c:1588): Restoring FAILED. 

That's because inode numbers (and thus -- file handles) has changed while you
copied the files. There's the --force-irmap option to criu dump/pre-dump
actions, that forces paths resolution for such cases. This option, in particular,
is used by p.haul when it live-migrates openvz container.

Thanks,
Pavel