13813 FIX api: addition of cache_ttl threshold to the metadata queue

The previous version caused the cluster collector to keep previous
metadata information of the colletors even if they were no longer
present within the cluster. This led to confusion on the handler
side as collectors appeared as still being present. This werk
fixes this behaviour and reduces the metadata retention time from
one year to the configured ttl cache time.

Change-Id: I4c86edd350a74707dffc6bf63773b8745f582d6c

.werks/13813

@@ -0,0 +1,15 @@
+Title: api: addition of cache_ttl threshold to the metadata queue
+Class: fix
+Compatible: compat
+Component: cluster-collector
+Date: 1646916855
+Knowledge: doc
+Level: 1
+Version: 2.0.0-alpha.1
+
+The previous version caused the cluster collector to keep previous
+metadata information of the colletors even if they were no longer
+present within the cluster. This led to confusion on the handler
+side as collectors appeared as still being present. This werk
+fixes this behaviour and reduces the metadata retention time from
+one year to the configured ttl cache time.

src/checkmk_kube_agent/api.py

@@ -334,6 +334,7 @@ def _init_app_state(
         key=metadata_key,
         maxsize=10000,  # Kubernetes clusters can have a max of 5000 nodes.
         # Each node collector daemonset sends their own metadata to this queue
+        ttl=cache_ttl,
     )
     app_.state.metadata = metadata
     app_.state.reader_whitelist = frozenset(reader_whitelist)