Mike Gerdts wrote:
> On 1/16/07, Dave Miner <Dave.Miner at sun.com> wrote:
>> Jacob Ritorto wrote:
>> > HI, This is only tangentially related to the forum topic, but no
>> > others seem to be discussing flar, so here goes: I want to use flar
>> > to do both speedy DR and regular backup for some T2000 machines.   I
>> > made a flar of the machine in its initial deployment state and wanted
>> > to just do diff flars from then on, which would be small and easy to
>> > store many on a tape.  When I tried to do the diff flar, it was
>> > unable to read the original flar (using the -A argument to
>> > flarcreate).  Turns out I have to have the image unbundled in its own
>> > directory for flarcreate to use it to compare.  This effectively
>> > kills my simple plan.  Any chance of getting flarcreate to be able to
>> > read .flar files instead of full directory heirarchies?  Any simple
>> > workaround that you know of?  I have no spare machines or disks to
>> > devote to the cloning and lu methods mentioned in the docs, and
>> > that's a big waste of space and time anyway..
> 
> It seems as though a good approach would be to attach a bart manifest
> as a section of the flash archive.  This would make it so that it
> would not need to read the entire flar to perform the comparison.
> 

Yeah, that would seem like an interesting optimization.

>> You certainly have the option to log an RFE through service, but it's
>> unlikely we'd work on it within the next year based on other priorities.
> 
> Based upon the fact that the flash archive utilities are not slated
> for being open sourced and new installation mechanisms are under
> active development, does it even make sense to hold out hope for such
> improvements?  Will the Caiman installer even use flash archives?
> (I've had a hard time figuring out just how much it will be
> replacing.)
> 

Fair questions.  Even if we change Flash, it's not impossible that such 
an enhancement would occur, since the Caiman work is targeted for a 
minor release after Solaris 10, and we could still enhance Solaris 10 
features independently.

However, we haven't decided anything about what flash will look like in 
Caiman; to be honest, we haven't thought about that issue much yet.  The 
issues around open-sourcing the current implementation would not prevent 
us from using and extending the existing archive formats, as they're 
specified independently and thus we can just construct a new 
implementation to the same formats.  We need to discuss the 
requirements, though; the disaster recovery and backup uses that people 
have found interesting weren't the original design so there may be 
reasons to take a different approach.

Overall, I think it's possible we can help meet this use case, but the 
time frame would be about what I originally stated.

Dave

Reply via email to