init

monitoring/ingest.py

@@ -0,0 +1,39 @@
+from django.db import transaction
+from django.utils import timezone
+from .models import Server,Event,Metric,Worker,Task
+STATES={'task-sent':'SENT','task-received':'RECEIVED','task-started':'STARTED','task-succeeded':'SUCCESS','task-failed':'FAILURE','task-retried':'RETRY','task-revoked':'REVOKED','task-rejected':'REJECTED'}
+TERMINAL={'SUCCESS','FAILURE','REVOKED','REJECTED'}
+@transaction.atomic
+def ingest(server_id,events):
+    server=Server.objects.select_for_update().get(pk=server_id)
+    count=0
+    for item in events:
+        event,created=Event.objects.get_or_create(id=item['id'],defaults={'server':server,'kind':item['kind'],'occurred_at':item['occurred_at'],'payload':item['payload']})
+        if not created:
+            if event.server_id!=server.id: raise ValueError('Event-ID gehört zu einem anderen Agenten')
+            continue
+        count+=1
+        p=event.payload; at=event.occurred_at
+        if event.kind=='metric':
+            Metric.objects.create(event=event,server=server,at=at,cpu=p['cpu'],ram=p['ram'],disk=p['disk'])
+            if not server.latest_sample_at or at>=server.latest_sample_at:
+                server.latest_sample_at=at; server.latest=p
+        elif event.kind.startswith('worker-'):
+            w,new=Worker.objects.get_or_create(server=server,hostname=p['hostname'],defaults={'last_seen':at})
+            if at>=w.last_seen:
+                w.last_seen=at; w.online=event.kind!='worker-offline'; w.active=p.get('active',0); w.processed=p.get('processed',0); w.save()
+        elif event.kind in STATES:
+            state=STATES[event.kind]
+            t,new=Task.objects.get_or_create(server=server,task_id=p['uuid'],defaults={'first_at':at,'last_at':at,'state':state})
+            t.first_at=min(t.first_at,at)
+            # A late RECEIVED must never overwrite a terminal result. Retry is nonterminal.
+            if (state in TERMINAL and t.state not in TERMINAL) or (at>=t.last_at and (t.state not in TERMINAL or state in TERMINAL)):
+                t.state=state; t.last_at=at
+                if p.get('runtime') is not None: t.runtime=p['runtime']
+                t.retries=max(t.retries,p.get('retries',0))
+            if p.get('name'): t.name=p['name']
+            if p.get('hostname'): t.worker=p['hostname']
+            t.save()
+    server.last_seen=timezone.now()
+    server.save(update_fields=['last_seen','latest_sample_at','latest'])
+    return count