Description of problem: I was trying to deploy 16.2 and the deployment failed due to a number of folders missing. So far I discovered /etc/pki/CA/*, /var/lib/pcsd, /var/log/pcsd, /var/lib/corosync, /var/log/cluster, /var/lib/ipa-client*. I am not sure if more are missing. (undercloud) [stack@undercloud-0 t]$ virt-customize -a overcloud-full.qcow2 --run-command 'rpm -V pcs ipa-client ipa-client-common openssl-perl pcs corosync' [ 0.0] Examining the guest ... [ 9.5] Setting a random seed [ 9.5] Setting the machine ID in /etc/machine-id [ 9.5] Running: rpm -V pcs ipa-client ipa-client-common openssl-perl pcs corosync missing /usr/lib/pcsd/vendor/bundle/ruby/doc missing /var/lib/pcsd missing /var/log/pcsd missing /etc/ipa/nssdb missing /var/lib/ipa-client/pki missing /var/lib/ipa-client/sysrestore missing /etc/pki/CA/certs missing /etc/pki/CA/crl missing /etc/pki/CA/newcerts missing /etc/pki/CA/private missing /usr/lib/pcsd/vendor/bundle/ruby/doc missing /var/lib/pcsd missing /var/log/pcsd missing /etc/corosync/uidgid.d missing /var/lib/corosync missing /var/log/cluster Those missing dirs make the deployment fail in different ways. Version: rhosp-director-images-x86_64-16.2-20200831.1.el8ost.noarch Puddle: http://download-node-02.eng.bos.redhat.com/rcm-guest/puddles/OpenStack/16.2-RHEL-8/RHOS_TRUNK-16.2-RHEL-8-20200831.n.0/ Workaround: In a pre config step do: dnf reinstall -y ipa-client ipa-client-common openssl-perl pcs corosync
This has ended up with me in a dib bucket; but from reading it appears to be that it has been root caused to a tmpfile daemon that has been clearing up files in the background on the build host? I'm not sure what dib can do about it? One idea is to not use the global /tmp but set DIB_TMP to some scratch location. We usually do this anyway to avoid blowing out tmpfs.
I think the internal image build scripts should be updated.. not a product bug. Perhaps a doc bug?
Did we ever make sure this completely went away? Because when I was looking into it previously, it still happened after doing the /tmp shuffle. This only started happening with RHEL8
Also I was wondering if this has the same root cause as Bug 1784001 which was still a thing unless something changed?
So i'll make the change to the rhosp-director-images distgit to specify DIB_TMP and we'll see if that resolves this for 16.2. Once we confirm that it resolves this we can apply to other versions.
IMHO this still needs to be updated in docs.. as customers may rebuild their images
We can look to add the DIB_TMP into the cli command instead (if not set).
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (Red Hat OpenStack Platform 16.2 deployment images), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHEA-2021:3485