Feature/hpc checks

checks.yaml

@@ -80,3 +80,18 @@
       host: https://homeserver.local
       ssc: True
       expected_code: 200
+
+- title: 'Group 4 / HPC'
+  checks:
+    - name: Slurm HPC Cluster
+      type: slurm
+      system: viper
+      expected: up
+      path: "../hpc/generic.json"
+      key_to_system: "systems"
+
+    - name: HPC Cluster Nodes
+      type: nodes
+      system: viper
+      path: "../hpc/generic.json"
+      key_to_system: "systems"

index.html.theme

@@ -187,6 +187,13 @@
                 <a target="_blank" href="{{ check.url }}">
             {% endif %}
         <div class="status-item">
+        {% if check.nodes %}
+                    <h3>{{ check.name }}</h3>
+                    {% for k, v in check.nodes.items() %}
+                        {{ k }}: {{ v }}<br />
+                    {% endfor %}
+                </div>
+            {% else %}
             <h3>{{ check.name }}</h3>
             <p class="{% if check.status %}status-up{% else %}status-down{% endif %}">
                 {{ '<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 256 256"><path d="M225.86,102.82c-3.77-3.94-7.67-8-9.14-11.57-1.36-3.27-1.44-8.69-1.52-13.94-.15-9.76-.31-20.82-8-28.51s-18.75-7.85-28.51-8c-5.25-.08-10.67-.16-13.94-1.52-3.56-1.47-7.63-5.37-11.57-9.14C146.28,23.51,138.44,16,128,16s-18.27,7.51-25.18,14.14c-3.94,3.77-8,7.67-11.57,9.14C88,40.64,82.56,40.72,77.31,40.8c-9.76.15-20.82.31-28.51,8S41,67.55,40.8,77.31c-.08,5.25-.16,10.67-1.52,13.94-1.47,3.56-5.37,7.63-9.14,11.57C23.51,109.72,16,117.56,16,128s7.51,18.27,14.14,25.18c3.77,3.94,7.67,8,9.14,11.57,1.36,3.27,1.44,8.69,1.52,13.94.15,9.76.31,20.82,8,28.51s18.75,7.85,28.51,8c5.25.08,10.67.16,13.94,1.52,3.56,1.47,7.63,5.37,11.57,9.14C109.72,232.49,117.56,240,128,240s18.27-7.51,25.18-14.14c3.94-3.77,8-7.67,11.57-9.14,3.27-1.36,8.69-1.44,13.94-1.52,9.76-.15,20.82-.31,28.51-8s7.85-18.75,8-28.51c.08-5.25.16-10.67,1.52-13.94,1.47-3.56,5.37-7.63,9.14-11.57C232.49,146.28,240,138.44,240,128S232.49,109.73,225.86,102.82Zm-52.2,6.84-56,56a8,8,0,0,1-11.32,0l-24-24a8,8,0,0,1,11.32-11.32L112,148.69l50.34-50.35a8,8,0,0,1,11.32,11.32Z"></path></svg> Operational' if check.status else '<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 256 256"><path d="M128,24A104,104,0,1,0,232,128,104.11,104.11,0,0,0,128,24Zm37.66,130.34a8,8,0,0,1-11.32,11.32L128,139.31l-26.34,26.35a8,8,0,0,1-11.32-11.32L116.69,128,90.34,101.66a8,8,0,0,1,11.32-11.32L128,116.69l26.34-26.35a8,8,0,0,1,11.32,11.32L139.31,128Z"></path></svg> Down' }}

tinystatus.py

@@ -61,6 +61,38 @@ async def check_port(host, port):
         return True
     except:
         return False
+
+async def check_slurm(expected, system, path, key_to_system):
+    try:
+        with open(path, "r") as file:
+            data = json.load(file)
+        slurm = data[key_to_system][system]
+        if slurm["state"] == expected:
+            return True
+        else:
+            print(f"SLURM check failed: {slurm["state"]} != {expected}")
+            return False                            
+
+    except aiohttp.ClientError as err:
+        print("Network error during SLURM check: ", err)
+        return False
+    except Exception as err:
+        print("Unexpected error with ", err)
+        return False
+
+async def check_nodes(system, path, key_to_system):
+    try:
+        with open(path, "r") as file:
+            data = json.load(file)
+        nodes = data[key_to_system][system]["nodes"]
+        return nodes
+
+    except aiohttp.ClientError as err:
+        print("Network error during SLURM check: ", err)
+        return []
+    except Exception as err:
+        print("Unexpected error with", err)
+        return []
 
 
 async def run_checks(checks):
@@ -76,17 +108,23 @@ async def run_checks(checks):
             task = tg.create_task(
                 check_http(check['host'], check['expected_code'], selfcert) if check['type'] == 'http' else
                 check_ping(check['host']) if check['type'] == 'ping' else
-                check_port(check['host'], check['port']) if check['type'] == 'port' else None,
+                check_port(check['host'], check['port']) if check['type'] == 'port' else 
+                check_slurm(check['expected'], check['system'], check['path'], check['key_to_system']) if check['type'] == 'slurm' else 
+                check_nodes(check['system'], check['path'], check['key_to_system']) if check['type'] == 'nodes' else None,
                 name=check['name']
             )
             if task:
                 background_tasks[check['name']] = task
 
+
     results = [
         {
             "name": check["name"],
             "url": check.get("url"),
-            "status": background_tasks[check["name"]].result()}
+            "status": background_tasks[check["name"]].result(),
+            **({"nodes": await check_nodes(check["system"], check["path"], check["key_to_system"]) }
+                if check["type"] == "nodes" else {})
+        }
         for check in checks
     ]