[sheepdog-users] [sheepdog] Impossible to convert a qcow2 to a vdi

[Hitoshi Mitake]

At Mon, 11 Aug 2014 11:59:28 +0800,
Liu Yuan wrote:
> 
> On Mon, Aug 11, 2014 at 10:54:46AM +0900, Hitoshi Mitake wrote:
> > At Fri, 8 Aug 2014 16:28:22 +0200,
> > Valerio Pachera wrote:
> > > 
> > > [1  <multipart/alternative (7bit)>]
> > > [1.1  <text/plain; UTF-8 (7bit)>]
> > > Sheepdog daemon version 0.8.0_293_g56a9dea
> > > 
> > > qemu-img convert -f qcow2 squeeze.qcow2 sheepdog:test
> > > qemu-img: cannot get vdi info, VDI isn't locked, test 0
> > > qemu-img: Could not open 'sheepdog:test': Input/output error
> > > 
> > > 
> > > sheep.log
> > > 
> > > Aug 08 16:22:28   INFO [main] md_add_disk(343) /mnt/sheep/0, vdisk nr 220,
> > > total disk 1
> > > Aug 08 16:22:28 NOTICE [main] get_local_addr(522) found IPv4 address
> > > Aug 08 16:22:28   INFO [main] send_join_request(982) IPv4 ip:192.168.10.5
> > > port:7000
> > > Aug 08 16:22:28 NOTICE [main] nfs_init(607) nfs server service is not
> > > compiled
> > > Aug 08 16:22:28   INFO [main] check_host_env(493) Allowed open files
> > > 1024000, suggested 6144000
> > > Aug 08 16:22:28   INFO [main] main(944) sheepdog daemon (version
> > > 0.8.0_293_g56a9dea) started
> > > Aug 08 16:23:44   INFO [main] rx_main(830) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35162, op=NEW_VDI, data=(not string)
> > > Aug 08 16:23:44   INFO [main] tx_main(882) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35162, op=NEW_VDI, result=16
> > > Aug 08 16:23:56   INFO [main] rx_main(830) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35164, op=MAKE_FS, data=(not string)
> > > Aug 08 16:23:56   INFO [main] tx_main(882) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35164, op=MAKE_FS, result=00
> > > Aug 08 16:23:58   INFO [main] rx_main(830) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35165, op=NEW_VDI, data=(not string)
> > > Aug 08 16:23:58   INFO [main] tx_main(882) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35165, op=NEW_VDI, result=00
> > > Aug 08 16:23:58   INFO [main] cluster_lock_vdi_main(1347) node: IPv4
> > > ip:192.168.10.5 port:7000 is locking VDI (type: shared): 7c2b25
> > > Aug 08 16:23:58   CRIT [main] vdi_lock(440) unknown type of locking: 0
> > > Aug 08 16:23:58  ERROR [main] cluster_lock_vdi_main(1350) locking
> > > 7c2b25failed
> > > Aug 08 16:24:04   INFO [main] rx_main(830) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35170, op=MAKE_FS, data=(not string)
> > > Aug 08 16:24:04   INFO [main] tx_main(882) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35170, op=MAKE_FS, result=00
> > > Aug 08 16:24:06   INFO [main] rx_main(830) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35171, op=NEW_VDI, data=(not string)
> > > Aug 08 16:24:06   INFO [main] tx_main(882) req=0x3705d50, fd=23, client=
> > > 127.0.0.1:35171, op=NEW_VDI, result=00
> > > Aug 08 16:24:07   INFO [main] cluster_lock_vdi_main(1347) node: IPv4
> > > ip:192.168.10.5 port:7000 is locking VDI (type: shared): 7c2b25
> > > Aug 08 16:24:07   CRIT [main] vdi_lock(440) unknown type of locking: 0
> > > Aug 08 16:24:07  ERROR [main] cluster_lock_vdi_main(1350) locking
> > > 7c2b25failed
> > > 
> > > Do I open a bug in launchpad?
> > 
> > This problem will be solved in the master branch of QEMU soon. You
> > don't have to open a bug in launchpad.
> > 
> 
> Unfortuntely, we can't fix the already distributed QEMU, this is why we try our
> best to keep backward compatibility and at the cost of ugly code sometimes.
> 
> Image that users of ubuntu, debian, fedora once update the sheepdog binary, then
> it can't work anymore, this would be huge compliaints. You can't expect all the
> users to update QEMU binary because
> 
> - many distributions maintain a stable branch, which consider the backward
>   compatiblity in the first place. So you'll never expect a QEMU master
>   break-compatilibty-patch backport to their stable branch.
> 
> - update QEMU means to stop running VM. This would be unacceptable for service
>   provider.

Nobody can promise 100% availability of services. Taking care about
stupid service providers who don't consider about software update is
meaningless.

> 
> So let's fix sheep. So make '0' as a lock type and grab the lock for '0' to keep
> backward compatibility.

OK, I'll update as you say because I really dislike wasting time on
this sort of discussion. I'm busy.

Thanks,
Hitoshi