Bug 1636493

This isn't a supported operation. You have to select cri-o at install time in the installer. From the comments it looks like cri-o was installed afterwards. Can you confirm that?

Created attachment 1492116 [details]
ansible inventory

No.  Cri-o was the runtime from the start/install.  Hosts file attached.

"bridge" is not supposed to be in /opt/cni/bin but it is installed in /usr/libexec/cni as part of the containernetworking-plugins rpm.

It should not be used on OpenShift.  What files do you have under /etc/cni/net.d/?  There should be only 80-openshift-network.conf and openshift-sdn.conf

I am not sure the installer should delete other configuration files if they are already present.

Here's what I have:

# ls -lh /etc/cni/net.d/
total 20K
-rw-r--r--. 1 root root 294 Oct 26 12:10 100-crio-bridge.conf
-rw-r--r--. 1 root root  54 Oct 26 12:10 200-loopback.conf
-rw-r--r--. 1 root root  83 Dec 10 09:57 80-openshift-network.conf
-rw-r--r--. 1 root root 483 Nov 15 23:01 87-podman-bridge.conflist
-rw-r--r--. 1 root root  82 Oct 11 17:03 openshift-sdn.conf

The openshift uninstaller does not remove any of the files. Removing them all per https://access.redhat.com/solutions/3449671 did resolve the issue after a full uninstall.

I'm seeing the same error when trying to setup local volumes[1] on 3.11 running cri-o (from install).

# ls -lh /etc/cni/net.d
total 16K
-rw-r--r--. 1 root root 294 Nov 15 20:45 100-crio-bridge.conf
-rw-r--r--. 1 root root  54 Nov 15 20:45 200-loopback.conf
-rw-r--r--. 1 root root  83 Dec  5 13:46 80-openshift-network.conf
-rw-r--r--. 1 root root  82 Dec  1 10:19 openshift-sdn.conf

Based on comments above and the KB article I'm not sure what is the proper resolution is.
Happy to provide inventory file if that's helpful.

[1] https://docs.openshift.com/container-platform/3.11/install_config/configuring_local.html

Just wanted to follow up and and say that my issue was resolved after deleting all the files but 80-openshift-network.conf and restarting the nodes.

Hello, it seems that this keeps happening after normal installs.
Would it be possible to add a check to the installer for these extra files, or fix the cases where it's not correctly removing the files?
Can we have an update on the status of the bug?

Thank you.

Gerard.

@gerodrig, there is a KB article explaining what to do in case of the issue.  I am not sure the installer should delete these files that were created separately.

I think the KB article is enough, Scott what do you think?

Without clearly documented reproducer steps I don't know if we should do anything more. Even with all the additional comments in this bug it's still not clear to me what the workflow is to produce this problem. If someone is doing something extra outside of the installer then I'm not worried about it, but if a simple installation to clean hosts according to our documented process results in a broken install we should fix that in the installer.

@(In reply to Mrunal Patel from comment #1)
> This isn't a supported operation. You have to select cri-o at install time
> in the installer. From the comments it looks like cri-o was installed
> afterwards. Can you confirm that?

Hello Mrunal,

Can you confirm that installing cri-o via rpm either *before* or after the installation is an unsupported operation?

Looking at the documentation [1] it's not clear that the only way to install cri-o should be letting the installation do it via the inventory variables.


[1] https://access.redhat.com/documentation/en-us/openshift_container_platform/3.11/html-single/cri-o_runtime/

CRI-O is only supported through the installer to guarantee that the nodes are in a good state and don't need any further manual steps.

(In reply to Mrunal Patel from comment #13)
> CRI-O is only supported through the installer to guarantee that the nodes
> are in a good state and don't need any further manual steps.

Thank you very much for the confirmation.

I've opened a documentation bug to make it clearer https://bugzilla.redhat.com/show_bug.cgi?id=1679130

(In reply to Mrunal Patel from comment #13)
> CRI-O is only supported through the installer to guarantee that the nodes
> are in a good state and don't need any further manual steps.

So how to deal this bug from the qe side since it's on ON_QA status?

I am not sure how to proceed here, the last comment doesn't specify what files needed to be deleted, but that is the solution provided in the KB.

The account who reopened the issue is disabled now so I cannot ask for more information.

Please reopen if it is still an issue.