[OpenAFS] CopyOnWrite failed: Partition ... may be out of free
inodes(errno = 2)
Tue, 19 Sep 2006 19:22:48 -0400
On Tuesday, September 19, 2006 03:15:25 PM -0700 Alf Wachsmann
> On Tue, 19 Sep 2006, Derrick J Brashear wrote:
>> On Tue, 19 Sep 2006, Alf Wachsmann wrote:
>> > Is this considered a bug and will be fixed eventually or is it a
>> > "feature" of the namei file server?
>> I consider it a bug but it's only really of interest to use the RO on a
>> different partition than RW when you have only one server in a cell, and
>> only then if you have multiple disks in the server. It's not really on
>> top of the list of usage cases.
> My concern here is not planned and deliberate action but accidents:
> I need to evacuate RW volumes from server2 onto server1 which already
> has a RO copy of some volumes somewhere. I overlook those RO volumes and
> move the RW volumes onto some other partition of that server.
> What if server1 now needs to be rebooted and the volumes need to be
> attached during the startup process. This will fail according to your
> earlier mail. Am I screwed now?
No. According to what Derrick observed, those volumes on the first
partition will be attached; any volumes in that group on other partitions
will not be. The workaround would be to remove the RO volumes and
rerelease them onto the same partition as the RW.
I agree with Derrick here - it's definitely a bug, but not currently a very
high priority one. I'm also not convinced it would be easy to fix. Even
if the bug were fixed, splitting RW and RO volumes from the same group
across partitions in this way would still not be good operational policy
(notwithstanding accidents like the scenario you describe).