Merge pull request #19356 from vespa-engine/freva/allow-more-inplace

Allow in-place resize for non-content nodes when decreasing resources…

node-repository/src/main/java/com/yahoo/vespa/hosted/provision/provisioning/NodePrioritizer.java

@@ -191,6 +191,7 @@ private NodeCandidate candidateFrom(Node node, boolean isSurplus) {
                                              parent.exclusiveToApplicationId().isEmpty()
                                              && requestedNodes.canResize(node.resources(),
                                                                          capacity.availableCapacityOf(parent),
+                                                                         clusterSpec.type(),
                                                                          topologyChange,
                                                                          currentClusterSize));
         } else {

node-repository/src/main/java/com/yahoo/vespa/hosted/provision/provisioning/NodeSpec.java

@@ -1,6 +1,7 @@
 // Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package com.yahoo.vespa.hosted.provision.provisioning;
 
+import com.yahoo.config.provision.ClusterSpec;
 import com.yahoo.config.provision.Flavor;
 import com.yahoo.config.provision.NodeFlavors;
 import com.yahoo.config.provision.NodeResources;
@@ -72,7 +73,7 @@ default boolean fulfilledBy(int count) {
      * in-place to resources in this spec.
      */
     default boolean canResize(NodeResources currentNodeResources, NodeResources currentSpareHostResources,
-                              boolean hasTopologyChange, int currentClusterSize) {
+                              ClusterSpec.Type type, boolean hasTopologyChange, int currentClusterSize) {
         return false;
     }
 
@@ -153,12 +154,12 @@ public boolean needsResize(Node node) {
 
         @Override
         public boolean canResize(NodeResources currentNodeResources, NodeResources currentSpareHostResources,
-                                 boolean hasTopologyChange, int currentClusterSize) {
+                                 ClusterSpec.Type type, boolean hasTopologyChange, int currentClusterSize) {
             // Never allow in-place resize when also changing topology or decreasing cluster size
             if (hasTopologyChange || count < currentClusterSize) return false;
 
-            // Do not allow increasing cluster size and decreasing node resources at the same time
-            if (count > currentClusterSize && !requestedNodeResources.satisfies(currentNodeResources.justNumbers()))
+            // Do not allow increasing cluster size and decreasing node resources at the same time for content nodes
+            if (type.isContent() && count > currentClusterSize && !requestedNodeResources.satisfies(currentNodeResources.justNumbers()))
                 return false;
 
             // Otherwise, allowed as long as the host can satisfy the new requested resources

node-repository/src/test/java/com/yahoo/vespa/hosted/provision/provisioning/InPlaceResizeProvisionTest.java

@@ -29,20 +29,17 @@
 import static org.junit.Assert.fail;
 
 /**
- * If there is no change in cluster size or topology, any increase in node resource allocation is fine as long as:
- * a. We have the necessary spare resources available on the all hosts used in the cluster
- * b. We have the necessary spare resources available on a subset of the hosts used in the cluster AND
- *    also have available capacity to migrate the remaining nodes to different hosts.
- * c. Any decrease in node resource allocation is fine.
+ * Node resources can be increased in-place if
+ *  1. No change to topology
+ *  2. No reduction to cluster size
+ *  3. There is enough spare capacity on host
  *
- * If there is an increase in cluster size, this can be combined with increase in resource allocations given there is
- * available resources and new nodes.
+ * Node resources can be decreased in-place if
+ *  1. No change to topology
+ *  2. No reduction to cluster size
+ *  3. For content/combined nodes: No increase to cluster size
  *
- * No other changes should be supported at this time, due to risks in complexity and possibly unknowns.
- * Specifically, the following is intentionally not supported by the above changes:
- *  a. Decrease in resource allocation combined with cluster size increase
- *  b. Change in resource allocation combined with cluster size reduction
- *  c. Change in resource allocation combined with cluster topology changes
+ * Node resources are increased if at least one of the components (vcpu, memory, disk, bandwidth) is increased.
  *
  * @author freva
  */
@@ -218,10 +215,10 @@ public void increase_size_decrease_resources() {
     public void cannot_inplace_decrease_resources_while_increasing_cluster_size() {
         addParentHosts(6, mediumResources.with(fast).with(local));
 
-        new PrepareHelper(tester, app).prepare(container1, 4, 1, mediumResources).activate();
-        assertSizeAndResources(container1, 4, new NodeResources(4, 8, 160, 1, fast, local));
+        new PrepareHelper(tester, app).prepare(content1, 4, 1, mediumResources).activate();
+        assertSizeAndResources(content1, 4, new NodeResources(4, 8, 160, 1, fast, local));
 
-        new PrepareHelper(tester, app).prepare(container1, 6, 1, smallResources);
+        new PrepareHelper(tester, app).prepare(content1, 6, 1, smallResources);
     }
 
     @Test(expected = OutOfCapacityException.class)