[sheepdog] the mounted vm of using sheepfs which open the object_cache became Read-only file system

周鹏 zhoupeng at cloudkc.net
Tue Aug 27 04:13:42 CEST 2013


>On Mon, Aug 26, 2013 at 06:40:49PM +0800, Liu Yuan wrote:
>> On Mon, Aug 26, 2013 at 09:47:41AM +0000, ?? wrote:
> >> >> It can know from the log that "volume_do_sync" will do CREATE_AND_WRITE_OBJ and then write something to the cache.
> > >>> In order to further confirmation, the source code is really to do so.
> > >>> But when and how did the vm became Read-only file system, i can not understand.So please help me to analyze the cause,thanks.
> >> 
> > >>If it is snapshot, it is read-only vdi.
> > >
> > >>> BTW,when i test the sheepdog, i often meet the vm became Read-only.So another question, what can cause a read_only?
> > >
> > >
> > >>Please use latest stable-0.7 or stable-0.6 branch and report problems if any.
> > >
> > >Yeah, i have updated the sheepdog to version 0.7.1_1_g66102e5.
>> > The same problem still exists.
>> > Under the path of "/home/centos6/root", i excute:
> >> 
>> > [root at sheep-dell root]# echo aa>test.txt
>> > -bash: test.txt: Read-only file system
>> > 
>> > At the same time, sheepfs printed:
>> > volume_do_sync(306): failed to flush vdi 7c7833
>> > volume_do_sync(306): failed to flush vdi 7c7833
>> 
>> Could you show me 'dog vdi list'?

[root at sheep-dell root]# dog vdi list
  Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
  centos6      0   10 GB  1.7 GB  0.0 MB 2013-08-27 09:23   7c7833     3 

>> 
>> I guess you were mounting a read-only vdi. I have tried write something into a
>> writeable vdi (either clone or a normal vdi) with success.
>> 
>> > 
>> > gataway log:
>> > 
>> > Aug 26 17:33:47  DEBUG [main] client_handler(786) 1, rx 0, tx 3
>> > Aug 26 17:33:47  DEBUG [main] finish_rx(590) 48, 127.0.0.1:49869
>> > Aug 26 17:33:47  DEBUG [main] queue_request(347) WRITE_OBJ, 1
>> > Aug 26 17:33:47  DEBUG [gway 13163] do_process_work(1334) 3, 7c78330000029d, 1

>> 
>> The log didn't report anything wrong.

Yes,so i really do not know how it became Read_only file system.

>Oops, I noticed that there is a fix for latest sheepfs. Could you try latest
>master branch to see if problem fixed?

>Then we can fix stable-0.7 and stable-0.6

Yeah, i test the sheepdog of 0.7.50.
After the vm is mounted,read and write obj became ok on linux and win7:

volume_do_rw(220): read oid 605eca00000319, off 1044480, len 4096, size 4096
volume_do_rw(220): read oid 605eca0000002e, off 73728, len 4096, size 4096
volume_do_rw(220): read oid 605eca0000031a, off 1441792, len 4096, size 4096
volume_do_rw(220): write oid 605eca00000019, off 1228800, len 4096, size 4096
volume_do_rw(220): write oid 605eca00000100, off 757760, len 4096, size 4096
volume_do_rw(220): write oid 605eca00000319, off 1044480, len 4096, size 4096
volume_do_rw(220): write oid 605eca00000319, off 1572864, len 4096, size 4096
volume_do_rw(220): write oid 605eca0000031a, off 1441792, len 4096, size 4096

So, what is the problem of version 0.6 and 0.7? How did the vm became Read_only file system? thanks.



More information about the sheepdog mailing list