Clusters card on home page not tied to underlying API server

[adamhgriffith-uofu]

Steps to reproduce

1. Create a Portal instance where SLATE_API_ENDPOINT = 'https://api.dev.slateci.io' in portal.conf.
2. Log into the Portal instance.
3. Note the content of the Clusters card on the home page and how production clusters are listed.
4. Click Clusters on the left navigation.
5. Note the correct set of SLATE clusters registered with the development API server.

Suggested fix

Update the Clusters card on the home page to display a subset of the clusters registered with the underlying API server.

Screenshot

[adamhgriffith-uofu]

This is causing logs noise in the non-production Portal applications when viewing the Dashboard page.

ERROR HIT: 'reachable'
[2022-08-04 14:30:44,774] ERROR in views_error_handling: Thu Aug  4 14:30:44 2022 Traceback occurred:
  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1513, in full_dispatch_request
    rv = self.dispatch_request()
n  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1499, in dispatch_request
    return self.ensure_sync(self.view_functions[rule.endpoint])(**req.view_args)
n  File "/home/agriffith/repos/slate-portal/portal/decorators.py", line 34, in decorated_function
    return fn(*args, **kwargs)
n  File "portal/views/views_clusters.py", line 156, in get_cluster_status_xhr
    cluster_status = get_cluster_status(cluster_name)
n  File "portal/views/views_clusters.py", line 171, in get_cluster_status
    cluster_status = cluster_status_query['reachable']

Traceback completed
127.0.0.1 - - [04/Aug/2022 14:30:44] "GET /cluster-status-xhr/uutah-prod HTTP/1.1" 200 -
ERROR HIT: 'reachable'
[2022-08-04 14:30:44,786] ERROR in views_error_handling: Thu Aug  4 14:30:44 2022 Traceback occurred:
  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1513, in full_dispatch_request
    rv = self.dispatch_request()
n  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1499, in dispatch_request
    return self.ensure_sync(self.view_functions[rule.endpoint])(**req.view_args)
n  File "/home/agriffith/repos/slate-portal/portal/decorators.py", line 34, in decorated_function
    return fn(*args, **kwargs)
n  File "portal/views/views_clusters.py", line 156, in get_cluster_status_xhr
    cluster_status = get_cluster_status(cluster_name)
n  File "portal/views/views_clusters.py", line 171, in get_cluster_status
    cluster_status = cluster_status_query['reachable']

Traceback completed
127.0.0.1 - - [04/Aug/2022 14:30:44] "GET /cluster-status-xhr/umich-prod HTTP/1.1" 200 -
ERROR HIT: 'reachable'
[2022-08-04 14:30:44,800] ERROR in views_error_handling: Thu Aug  4 14:30:44 2022 Traceback occurred:
  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1513, in full_dispatch_request
    rv = self.dispatch_request()
n  File "/home/agriffith/.conda/envs/slate-portal/lib/python3.6/site-packages/flask/app.py", line 1499, in dispatch_request
    return self.ensure_sync(self.view_functions[rule.endpoint])(**req.view_args)
n  File "/home/agriffith/repos/slate-portal/portal/decorators.py", line 34, in decorated_function
    return fn(*args, **kwargs)
n  File "portal/views/views_clusters.py", line 156, in get_cluster_status_xhr
    cluster_status = get_cluster_status(cluster_name)
n  File "portal/views/views_clusters.py", line 171, in get_cluster_status
    cluster_status = cluster_status_query['reachable']

[adamhgriffith-uofu]

Doing more digging, in portal/views.py is the following:

if request.method == 'GET':
        print("Session from dashboard: {}".format(session))
        if session["slate_portal_user"]:
            # single-user mode
            clusters = ["my-cluster"]
        else:
            clusters = ["uutah-prod", "uchicago-prod", "umich-prod"]

        with open('portal/static/news.md', "r") as file:
            news = file.read()
        # This json conversion is for JS to read on the frontend
        clusters_list = json.dumps(clusters)

        return render_template('dashboard.html', clusters=clusters,
                                news=news, clusters_list=clusters_list)

Since clusters is hard-coded this causes issues with the Vue expression in portal/templates/dashboard.html:

new Vue({
	el: '#clusters-card',
	data () {
    return {
      clusters: {},
      loading: true,
      failed: false
    }
	},
	async mounted() {
    let status_results = [];
    await Promise.all(clusters.map(cluster =>
      axios.get('/cluster-status-xhr/' + cluster).then(response => {
        status_results.push([cluster, response.data]);
        })
      ));
    this.clusters = status_results;
		}
  });

where we discover an API call in portal/views/views_clusters.py to clusters that are not registered with the slate_api_endpoint:

@app.route('/cluster-status-xhr/<cluster_name>', methods=['GET'])
@authenticated
def get_cluster_status_xhr(cluster_name):
    """
    - List Clusters Registered on SLATE (json response)
    """
    if request.method == 'GET':
        cluster_status = get_cluster_status(cluster_name)
        # print(cluster_status)
        return jsonify(cluster_status)


def get_cluster_status(cluster_name):
    """
    - Get Clusters and Status on SLATE
    """
    access_token = get_user_access_token(session)
    query = {'token': access_token, 'cache': 'cache'}

    cluster_status_query = requests.get(
            slate_api_endpoint + '/v1alpha3/clusters/' + cluster_name + '/ping', params=query)
    cluster_status_query = cluster_status_query.json()
    cluster_status = cluster_status_query['reachable']
    
    return cluster_status

[adamhgriffith-uofu]

I could make the errors go away for right now without "truly" fixing the issue by doing the following in portal/views.py:

if session["slate_portal_user"]:
    # single-user mode
    clusters = ["my-cluster"]
elif slate_api_endpoint == "https://api.dev.slateci.io":
    clusters = ["river-dev-fp"]
elif slate_api_endpoint == "https://api.staging.slateci.io":
    clusters = ["uutah-dev"]
else:
    clusters = ["uutah-prod", "uchicago-prod", "umich-prod"]

[jstidd]

You're right that the clusters on the card should not be hardcoded, but which clusters do we show? The first 5 in the list? The first five may not be the ones we want to highlight. Maybe it can be split, if in production it shows the main three production sites, but in staging and dev it shows the first five. Just some thoughts.