One EKS cluster not clickable in the UI

[gaktive]

Internal reference: SURE-7035
Reported in 2.7.6

Issue description:
The cluster EKS eks-entsvcs-pci-dev (cluster ID: c-nqdv4) is not clickable from the explorer after Rancher upgrade from 2.6.12 to 2.7.6

Business impact:
The user can't click on that cluster from the explorer

Troubleshooting steps:
reviewed the rancher pods and cluster agent logs
rolling restart of rancher pods
re-deployed the cluster agents

Repro steps:
Support was not able to reproduce it, but the user wants to know what is causing this behavior

Workaround:
None

Using the cluster management or directly accessing the link for that cluster with something like this https://hostname/dashboard/c/c-xxxxx/explorer

Actual behavior:
A cluster is not clickable from the explorer

Expected behavior:
A cluster should be and clickable when it is Active in Rancher

Addenda:

• Logs & HAR files available via the SURE ID.

[realhuseyn]

Any updates? I have faced this issue

[aalves08]

Going through the code in 2.7.6 and checking the objects provided in the JSON and HAR files for both the management and provisioning clusters, the homepage and side menu links are disabled because the management object has a condition with error: true there which is needed for enabling the homepage link

Check for main nav link:
x.isReady && !pCluster?.hasError where x is of type management cluster => fails because prov cluster has a status.condition with error: true => side menu https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/provisioning.cattle.io.cluster.js#L863-L865. 🛑 fails on pCluster.hasError

Check for homepage link:
row.mgmt.isReady && !row.hasError where row is of type provisioning cluster => fails because prov cluster has a status.condition with error: true => side menu https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/provisioning.cattle.io.cluster.js#L863-L865. 🛑 fails on row.hasError

check for /explorer route direct access:
!cluster.isReady where cluster is of type management cluster => works because it has a Connected condition ✅

https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/management.cattle.io.cluster.js#L173-L181
https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/provisioning.cattle.io.cluster.js#L863-L865

The workaround (direct link to /explorer) works because the authenticated middleware is just checking for isReady on
the store action loadCluster for the management cluster https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/management.cattle.io.cluster.js#L173-L181, which in turn has the correct checks for it to work.

The question here is: which is correct set of condition for us to check for both links?

FYI @nwmac @richard-cox

[aalves08]

The explore button on Cluster Managament should also be disabled since:

row.mgmt && row.mgmt.isReady && !row.hasError where row is of type provisioning cluster=> fails because prov cluster has a status.condition with error: true => side menu https://github.com/rancher/dashboard/blob/release-2.7.6/shell/models/provisioning.cattle.io.cluster.js#L863-L865. 🛑 fails on row.hasError

[izaac]

@yonasberhe23 the test coverage looks good to me, the scenario is really hard to reproduce for a manual validation after going through the issue comments. We can move to done.