[OpenAFS] Re: Vos functions and clones and shadows

Steve Simmons scs@umich.edu
Tue, 26 Jun 2007 18:12:32 -0400

On Jun 26, 2007, at 5:41 PM, Adam Megacz wrote:

> Derrick J Brashear <shadow@dementia.org> writes:
>> the problem is it lives in the volume group of its parent, so it's a
>> halfling. parent's gone, but... yeah, like you say
> While we're on the topic, what is the reason for having volume groups
> in the first place rather than just having a reference in the volume
> header saying "I am a backup volume for volume XYZ" to establish
> volume relationships?

Oh man, Dan and I came up with all sorts of scenarios about what  
volume groups are. I believe that we ultimately decided that to make  
them really useful, you'd have to add some stuff to the vldb and that  
a great deal of parent/child volume tracking would have to be cleaned  
up in the current code. In the end we opted to do minimal change to  
the current code until/if there was a clearer purpose and use for them.

> Is the "volume numbers share all but the last three bits" criterion
> visible to the cache manager, or is this something that could be
> altered just on the servers and admin clients (vos, bos, etc)?

Good question. Given the possibility of hanging clones and the  
utility we see for having clones of clones (get around the 7-clone  
limit in namei, albeit at major cost of disk space), I don't see how  
one can avoid having to make the cache managers deal with it at some