[OpenAFS] restoring root.cell (RW) when main server dies....

Christopher Arnold chris@pictage.com
Fri, 25 Jul 2003 16:27:23 -0700


Just want to say that I took the safer first method since it was
root.cell and root.afs;
things went really smoothly. Thanks for the quick response on this one!  

Are there any suggestions for improving our design to keep things going
in the event
these important volumes go down?  For now I think I'm going to switch
things around
so that we've got a server or two that does nothing except serve up critical
volumes like these (we rely heavily on RW paths for certain apps.....).
 Currently the
RW volumes are scattered among machines which happen to do a tremendous
amount of I/O.


-c




Phil.Moore@morganstanley.com wrote:

>>>>>>"Derek" == Derek Atkins <warlord@MIT.EDU> writes:
>>>>>>            
>>>>>>
>
>Derek> vos dump | vos restore
>
>May I respecfully suggest that you *always* use an intermediate file
>for this kind of data migration?
>
>Yes, Derek, I know you're paraphrasing, but seriously, if you screw up
>the vos restore arguments, then vos dump dies with a SIGPIPE, and
>leave the volume transaction hanging.
>
>The timeout for those transactions is on the order of minutes, so if
>you are doing this en masse, or doing this in an emergency, that 5
>minute (or whatever it is) timeout can be painful.
>
>IOW...
>
>vos dump to a file
>vos restore from the file
>
>will usually be much safer.
>
>
>
>  
>

-- 
Christopher Arnold
System Administrator
Pictage, Inc.