Trying to fix failed sync

This change makes the rule a preference instead of a hard requirement, allowing Kubernetes to schedule the pod even if no nodes fully meet the affinity rule, but still trying to respect the affinity where possible.

Author: ranguard

apps/web/base/deployment.yaml

@@ -18,7 +18,7 @@ spec:
       # spread nodes across servers
       affinity:
         podAntiAffinity:
-          requiredDuringSchedulingIgnoredDuringExecution:
+          preferredDuringSchedulingIgnoredDuringExecution:
           - labelSelector:
               matchExpressions:
               - key: app