On Friday, March 24, 2006 08:04:42 PM -0800 Russ Allbery <[EMAIL PROTECTED]>
wrote:
We moved a pre-existing test volume from another server onto a 1.4.1-rc10
server earlier this week, and now vos examine output shows zeroed
timestamps for volume creation and last update:
user.wjt 2003745036 RW 41293 K On-line
afssvr13.Stanford.EDU /vicepa
RWrite 2003745036 ROnly 0 Backup 2003745038
MaxQuota 215000 K
Creation Wed Dec 31 16:00:00 1969
Copy Tue Mar 21 16:55:57 2006
Backup Fri Mar 24 02:15:01 2006
Last Update Never
RWrite: 2003745036 Backup: 2003745038
number of sites -> 1
server afssvr13.Stanford.EDU partition /vicepa RW Site
The same output shows up in vos listvol -long.
In Zephyr conversation, Derrick thought this may have been due to an
intentional change. I'd like to understand more about this and whether we
can prevent it from happening somehow. We have fairly extensive AFS
reports that rely, among other things, on those values to determine who is
actually using their AFS space and how recently they've used it, and we
need those values to be preserved across volume moves.
Note that this isn't a case, so far as I can tell, of simple latency in
updating that information. The volume was moved onto that server three
days ago.
Hm. This might be the result of an interation between the way vos decides
what values to print and a change in how moves work. In particular, the
zero creation time disturbs me somewhat.
Unfortunately, vos lies sometimes and doesn't always tell you what values
are really in some of those fields, so they can be a bit useless for
debugging.
Try examining that volume with -format to see the real values.
I bet you'll find that if you move a volume to one of your existing
servers, immediately after the move a normal vos examine will report the
same values for create, copy, and modify dates, but using -format will show
that the modify date is actually zero. Moving a volume to the new server
is probably resulting in both the create and modify dates being set to
zero, though I haven't yet determined why that is happening.
-- Jeff
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info
