Fix grafana login issue

[w13915984028]

Problem:

The grafana login page showed 404.

A main reason is, when accesing grafana from the local cluster, the general URL is https://1.2.3.4/k8s/clusters/local/api/v1/...., the Rancher will trim /k8s/clusters/local/, this works fine for normal link. But for login, it is another story, the grafana will return a 302 code with a new Location, unfortunately, the Rancher will NOT add /k8s/clusters/local back to the Location. When grafana's front page code runs, it will get a different appSubUrl and assume error, then showed 404 page.

For rancher managed cluster, the regex processing in appSubUrl is not perfect to deal with 302.

Solution:

(1) For local cluster, add a fallback item in the nginx sub-filter
(2) For rancher-managed cluster, use long match in regex to avoid duplicate URL elements.

Related Issue:
harvester/harvester#4756

Test plan:

1. Install a new Harvester v1.3.0 cluster, enable rancher-monitoring addon. Or use this workaround [BUG] Grafana (Addon: rancher-monitoring) sign in is not working in Harvester 1.2.1 harvester#4756 (comment) in existing v1.2.1/v1.3.0 cluster.
2. Click grafana from the Dashboard page, which will open a new page.
3. Click login, try login; or return to previous page, all should run smoothly.

In Rancher-integration scenario, repeat the same processes.

[bk201]

@w13915984028 If the user turns on the add-on in v1.2.1, does this fix work if he upgrades to the newer version?

[w13915984028]

@bk201

For v1.2.1 -> v1.3.0,
After harvester/harvester#4941, the rancher-monitoring-crd will have a newer version, the upgrade process will make sure the rancher-monitoring-crd managedchart is upgraded with new value, for all PODs in cattle-monitoring-system, they will be replaced when NODE is rebooted/addon is upgraded, the bug will be fixed after upgrade.

For v1.2.1 -> potential v1.2.2, we will not bump a newer version of rancher-monitoring chart, the fleet-agent may not be able to upgrade the rancher-monitoring-crd managedchart as the chart version is not changed, then the configmap is not updated. Tracked in harvester/harvester#4942.

[bk201]

@bk201

For v1.2.1 -> v1.3.0, After harvester/harvester#4940, the rancher-logging-crd will have a newer version, the upgrade process will make sure the rancher-logging-crd managedchart is upgraded with new value, for all PODs in cattle-logging-system, they will be replaced when NODE is rebooted/addon is upgraded, the bug will be fixed after upgrade.

For v1.2.1 -> potential v1.2.2, we will not bump a newer version of rancher-logging chart, the fleet-agent may not be able to upgrade the rancher-logging-crd managedchart as the chart version is not changed, then the configmap is not updated. Tracked in harvester/harvester#4756.

Thanks for the explanation, but did you mean the rancher-monitoring chart, not the logging one?

[w13915984028]

@bk201 You are right :) , it is rancher-monitoring not rancher-logging related, the description is updated.

[bk201]

LGTM, thanks!

[Vicente-Cheng]

LGTM!

[bk201]

@Mergifyio backport v1.2

[mergify]

backport v1.2

✅ Backports have been created

• #626 Fix grafana login issue (backport #623) has been created for branch v1.2