[Info-vax] Uptime for OpenVMS

[Johnny Billquist]

On 2011-05-13 01.32, Bob Eager wrote:
> On Fri, 13 May 2011 00:54:09 -0600, Johnny Billquist wrote:
>
>> On 2011-05-13 00.01, Bob Eager wrote:
>>> On Fri, 13 May 2011 01:21:48 +0000, glen herrmannsfeldt wrote:
>>>
>>>> Steve Thompson<smt at vgersoft.com>   wrote:
>>>>
>>>> (snip)
>>>>> I have a lot of Linux NFS servers and a lot of NFS clients, mostly
>>>>> Linux and some OSX, and over several years I have never seen a server
>>>>> reboot cause any problems for a client, not that a server reboot
>>>>> happens often.
>>>>
>>>> It is only supposed to fail (stale file handle) if, for example, you
>>>> replace the disk drive with a completely different one. You wouldn't
>>>> want data from a file open on the previous disk to be written to the
>>>> new one.
>>>
>>> Even then, if the inode numbers are preserved, it shpuld be OK.
>>
>> No. Because at that point, the inode numbers might mean totally
>> different files.
>
> Not what I said. I said "IF the inode numbers are preserved". I accept
> that this may not necessarily be the case.

Ok. True. Yes, if the inode numbers were preserved, then it would be 
theoretically possible.
However, the OS have no way of knowing for sure that the inode numbers 
were preserved, so that is more on the theoretical side (I actually 
don't know of any way to actually preserve inode numbers when copying a 
disk...).

>> A full file identification is a combination of the inode number, and a
>> device identifier.
>
> And the device identifier need not change.

No. But the OS should probably always do this, because it can't decide 
for sure, and taking a chance with this is a good way to disaster.

>> inode number themselfs are only guaranteed to be unique within one file
>> system. Even on local disks, you can have the same inode number for
>> different files, if they are on different local disks.
>
> Yes, I know that. In fact, the same local disk can include the same inode
> number multiple times, since the disk is likely to be partitioned into
> distinct file systems.

Right.

	Johnny