[cgl_discussion] rpm rollback failed transaction patch...

James Olin Oden joden at malachi.lee.k12.nc.us
Tue Jun 3 10:59:01 PDT 2003

On 3 Jun 2003, Mika Kukkonen wrote:

> Hello James!
> Sorry about our lack of response, we have been quite busy in finalizing
> our v2.0 spec. Anyway, I looked at your web site (link below) and I
> think the concept looked good.
I completely understand being busy.  Its not a problem.  BTW, I have found
a problem with my patch and I am trying to fix that today.  You will 
probably see another email concerning the new patch within the next 
day after I have it all tested.
> Concerning our specs, we do have a requirement for remote upgrade
> automatic rollback, and we'll add your web page as a PoC referral
> to that one. Thanks!
I am highly interested with the remote upgrade and the live upgrade 
that I believe I saw listed.  Let me know if there is anyway I can help.
Presently, we write our own upgrade, and backout scripts around rpm, 
but the problem is general enough that surely a mechanism could be written
that all parties could use (including the company for which I work).  
Unfortunately, its really hard to get the some of the RedHat folks and 
others writting update mechanisms that communicating with a remote system
across the internet (or otherwise) is not an option in most cases in a CG
environment.  up2date actually could make a good starting point if one 
could just get it to look of its repository of rpms off a CDROM, and to 
make its decisions without talking to remote server.  Of course, then 
there is the issue of creating a chrooted environment such that all tools
used to perform the upgrade (rpm for instance) are not comming from the 
local system, along with on systems tht are mirrored using an approach
of breaking the mirrors and upgrading the broken off mirror.  

Anyway, there is a lot of work to do.


More information about the cgl_discussion mailing list