Created attachment 1477315 [details] inventory file used for CRS. Description of problem: ------------------------------- The independent mode of deployment threw error while running the prerequisite ansible playbook. It is searching for the python-docker package on the gluster nodes. The gluster nodes should neither have nor look for the ocp related packages because gluster nodes are out of CNS setup. Version-Release number of selected component: -------------------------------- oc v3.10.14 glusterfs-3.12.2-16 How reproducible: ------------------------------ Everytime Steps to Reproduce: ------------------------------- 1.Install Red Hat Gluster Storage Server on RHEL. 2.Configure port access,kernel module and enable sshd,gluster-blockd and glusterd services. 3.Edit the CRS inventory file and run the ansible playbook for the prerequisites. Actual results: ------------------------------ It threw an error looking for python docker package in gluster nodes. Expected results: ----------------------------- Python-docker package is not required in gluster nodes.
Created attachment 1477395 [details] ansible log file captured on the master
## In reply to comment 12 by Jose A.Rivera. Jose , i tried deployment with suggested changes but it gave a new error. I have attached the ansible logs and setup details where i faced this issue while deploying metrics,registry and logging.
Created attachment 1478400 [details] enterprise error
I have also seen the similar problem with my setup. Here is the snip of the console log. <snip> TASK [Run variable sanity checks] ********************************************************************************************************************************************************************************* fatal: [dhcp47-157.lab.eng.blr.redhat.com]: FAILED! => {"failed": true, "msg": "last_checked_host: dhcp47-22.lab.eng.blr.redhat.com, last_checked_var: openshift_deployment_type;openshift_deployment_type must be defined and one of origin, openshift-enterprise"} to retry, use: --limit @/usr/share/ansible/openshift-ansible/playbooks/prerequisites.retry PLAY RECAP ******************************************************************************************************************************************************************************************************** dhcp46-205.lab.eng.blr.redhat.com : ok=14 changed=0 unreachable=0 failed=1 dhcp46-224.lab.eng.blr.redhat.com : ok=31 changed=6 unreachable=0 failed=0 dhcp46-51.lab.eng.blr.redhat.com : ok=31 changed=6 unreachable=0 failed=0 dhcp47-145.lab.eng.blr.redhat.com : ok=31 changed=6 unreachable=0 failed=0 dhcp47-157.lab.eng.blr.redhat.com : ok=49 changed=6 unreachable=0 failed=1 dhcp47-2.lab.eng.blr.redhat.com : ok=14 changed=0 unreachable=0 failed=1 dhcp47-22.lab.eng.blr.redhat.com : ok=15 changed=0 unreachable=0 failed=1 localhost : ok=14 changed=0 unreachable=0 failed=0 </snip>
I'm going to guess that your inventory file has changed, since the following machines were not listed in the one you posted at the beginning: [dhcp46-221.lab.eng.blr.redhat.com] [dhcp46-94.lab.eng.blr.redhat.com] [dhcp46-106.lab.eng.blr.redhat.com] In any case, it seems I was wrong! Please put back the glusterfs group names under [OSEv3:children]. Try manually installing python-docker on all the CRS nodes and see if that fixes the problem. We'll figure out what the right fix is afterwards.
Created attachment 1478592 [details] inventory file with glusterfs removed from OSEv3:children
Created attachment 1478593 [details] ansible.log file corresponding to execution with inventory file without gluster in OSEv3:children
Created attachment 1479042 [details] Change g_all_hosts patch I'm cooking up a potential fix, but I don't currently have an environment where I can test an external GlusterFS cluster that's not on OCP nodes. If someone (Mugdha?) is willing to experiment for me, would you be willing to apply the following patch (with "patch" or manually) to the files in /usr/share/ansible/openshift-ansible? If you download the patch to the home directory, you can apply it with the following: patch -p1 -d /usr/share/ansible/openshift-ansible/ <~/groups.patch To remove it (in case it breaks everything), just do: patch -p1 -d /usr/share/ansible/openshift-ansible/ -R <~/groups.patch
Oops, shouldn't have checked the patch box. :) URL to the raw patch: https://bugzilla.redhat.com/attachment.cgi?id=1479042&action=diff&context=patch&collapsed=&headers=1&format=raw
On consolidation from comment24 and comment25, there are 5 packages that are required to be made available with RHGS nodes to support independent mode of OCS installation: - python-docker ( ose-3.10 repo ) - python-docker-pycreds ( extras repo ) - python-websocket-client ( extras repo ) - python2-urllib3 ( ose-3.10 repo ) - python2-pysocks ( ose-3.10 repo )
This PR may help solve the issue: https://github.com/openshift/openshift-ansible/pull/9847
PR merged. Backport PR:https://github.com/openshift/openshift-ansible/pull/9934
Backport PR merged.
(In reply to Jose A. Rivera from comment #21) > Created attachment 1479042 [details] > Change g_all_hosts patch > > I'm cooking up a potential fix, but I don't currently have an environment > where I can test an external GlusterFS cluster that's not on OCP nodes. If > someone (Mugdha?) is willing to experiment for me, would you be willing to > apply the following patch (with "patch" or manually) to the files in > /usr/share/ansible/openshift-ansible? If you download the patch to the home > directory, you can apply it with the following: > > patch -p1 -d /usr/share/ansible/openshift-ansible/ <~/groups.patch > > To remove it (in case it breaks everything), just do: > > patch -p1 -d /usr/share/ansible/openshift-ansible/ -R <~/groups.patch Missed to update the bug. I have tested with this patch and the error is gone