Anyone willing to test a restore roll?

[Philip Papadopoulos]

I have a procedure that I would like  (willing) folks to test for a restore roll.

Background: 

        Yum is much more pedantic about directory ownership (which RPM owns what).

         This means that RPMs generated during the creation of restore roll on 6 (and 

         as of right now, on 7, too). Can't be installed/used.  Those RPMs need to

         be modified.

Process:  The latest core roll has two new scripts  (listisofiles, repack-roll).  If you download the latest core roll, you can add this to your current 7 cluster, rebuild the distro and do a yum update. You will get the updated rocks-admin package installed.

#1 create a restore roll in the usual way (on 6 or 7)

#2  install a new frontend WITHOUT the restore roll.

#3 copy the restore iso image that you created to your newly installed frontend

#4 repack-roll <name of isoimage>

     you will get a new isoimage of the same name with a "repacked" suffix.

#5 add the restore roll (the one with repacked suffix, enable it , rebuild the distro)

#6 rocks run roll <name of your restore roll> | sh

(you should get warnings/errors that "things" already exist, this is expected)

#7 fixup (one last step).

     cat /root/.ssh/id_rsa.pub >> /etc/ssh/authorized_keys/id_rsa.pub

At this point, your users should be restored, your database of nodes, networks, etc. should be restored.  (old attrs won't override existing ones).

When you kickstart nodes, the names of the interfaces will be renamed properly

(eg. eth0 --> enps0f0 type of naming)

Thanks,

Phil 

--

Philip Papadopoulos, Ph.D

ppapad...@ucsd.edu

(858) 822-3628