Ok, I’m trying way too hard to shovel myself out of a hole.

We have a bunch of remote CentOS 6 servers, that were configured with kickstart.

They’ve subsequently had additional RPMs installed/updated via the
‘rpm’ utility.

We have reason to occasionally rewind the state of the server back to it’s original set of RPMs.

Of late, we’ve found that ‘yum’ handles this scenario quite nicely, IFF (if and only if) that all RPM changes were handled via the ‘yum’
utility itself, such that yum can manage a list of transactions. In this case ‘yum history rollback 1’ does exactly what we need.

My problem is that we did not exclusively use ‘yum’ to do this work, and even though there are newer RPMs in place, yum only knows about transaction ID 1, from the initial kickstart installation. Hence, it thinks there’s nothing to rollback from.

So, given the knowledge of what RPMs are currently installed, is there a way to invent a new transaction incorporating them, without reinstalling the RPMs in question?

I feel I’ve come close, with this methodology, basically coercing yum to rebuild a database in a private directory:

# installroot for yum
workdir=/home/yum-hack

rm -rf ${workdir}
mkdir -p ${workdir}
cd ${workdir}

# list of newer RPMs (after transaction ID 1)
new=new_rpms

# dup repos and rpm/yum databases
for tree in /var/lib/rpm/ /var/cache/yum/ /var/lib/yum/ /etc/yum.repos.d/; do
echo ${tree}
mkdir -p ./${tree}
cp -pr ${tree}/* ./${tree}
done

cat new_rpm | xargs /usr/bin/yum -y –nogpgcheck –installroot=${workdir} \
–setopt=tsflags=noscripts reinstall

On the surface, this seems to work, in that I now can see a transaction list:

yum –installroot=/home/yum-hack history list all

But, any effort to rollback to tid 1 yields:

Transaction history is incomplete, before 5.
You can use ‘history rollback force’, to try anyway.
Error: Failed history rollback, incomplete

So, I’ve clearly tricked myself.

Does anyone have any suggestions about how I _should_ retroactively invent an additional yum transaction?

This there some state I failed to introduce into the the above private directory?

I’ve only begun to dive into the yum source to try to track down what triggers this error, but if anyone has some immediate knowledge, or suggestion for a better forum where I can ask this question?

Some details of my environment, if that’s helpful:

[root@172-20-245-145 ~]# cat /etc/redhat-release
CentOS release 6.5 (Final)
[root@172-20-245-145 ~]# rpm -q yum rpm
yum-3.2.29-69.el6.CentOS.noarch
rpm-4.8.0-37.el6.x86_64

Thanks for any advice…