Merge pull request #13085 from vespa-engine/gjoranv/reduce-#threads

Limit number of threads for applications with <20 nodes.

2 files changed, 23 insertions, 7 deletions

diff --git a/metrics-proxy/src/main/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetriever.java b/metrics-proxy/src/main/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetriever.java
index 4d57e89fc76..40011a0dc72 100644
--- a/metrics-proxy/src/main/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetriever.java
+++ b/metrics-proxy/src/main/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetriever.java
@@ -19,7 +19,6 @@ import java.util.List;
 import java.util.Map;
 import java.util.concurrent.ForkJoinPool;
 import java.util.concurrent.TimeUnit;
-import java.util.logging.Level;
 import java.util.logging.Logger;
 import java.util.stream.Collectors;
 
@@ -37,16 +36,16 @@ public class ApplicationMetricsRetriever extends AbstractComponent {
 
     private static final Logger log = Logger.getLogger(ApplicationMetricsRetriever.class.getName());
 
-    private static final int PARALLELISM = 20;
+    static final int MAX_THREADS = 20;
     static final Duration MIN_TIMEOUT = Duration.ofSeconds(60);
-    private static final Duration MAX_TIMEOUT = Duration.ofSeconds(240);
+    static final Duration MAX_TIMEOUT = Duration.ofSeconds(240);
 
     private static final int HTTP_CONNECT_TIMEOUT = 5000;
     private static final int HTTP_SOCKET_TIMEOUT = 30000;
 
     private final HttpClient httpClient = createHttpClient();
     private final List<NodeMetricsClient> clients;
-    private final ForkJoinPool forkJoinPool = new ForkJoinPool(PARALLELISM);
+    private final ForkJoinPool forkJoinPool;
 
     // Non-final for testing
     private Duration taskTimeout;
@@ -55,7 +54,9 @@ public class ApplicationMetricsRetriever extends AbstractComponent {
     @Inject
     public ApplicationMetricsRetriever(MetricsNodesConfig nodesConfig) {
         clients = createNodeClients(nodesConfig);
-        taskTimeout = timeout(clients.size());
+        int numThreads = Math.min(clients.size(), MAX_THREADS);
+        taskTimeout = timeout(clients.size(), numThreads);
+        forkJoinPool = new ForkJoinPool(numThreads);
     }
 
     @Override
@@ -112,8 +113,8 @@ public class ApplicationMetricsRetriever extends AbstractComponent {
                 .build();
     }
 
-    private static Duration timeout(int clients) {
-        Duration timeout = Duration.ofSeconds(Long.max(MIN_TIMEOUT.toSeconds(), 20 * clients / PARALLELISM));
+    static Duration timeout(int clients, int numThreads) {
+        Duration timeout = Duration.ofSeconds(Long.max(MIN_TIMEOUT.toSeconds(), 20 * clients / numThreads));
         return timeout.compareTo(MAX_TIMEOUT) > 0 ? MAX_TIMEOUT : timeout;
     }
 
diff --git a/metrics-proxy/src/test/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetrieverTest.java b/metrics-proxy/src/test/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetrieverTest.java
index 1f2852e3526..a74989f7af6 100644
--- a/metrics-proxy/src/test/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetrieverTest.java
+++ b/metrics-proxy/src/test/java/ai/vespa/metricsproxy/http/application/ApplicationMetricsRetrieverTest.java
@@ -11,6 +11,10 @@ import java.util.Arrays;
 import java.util.concurrent.TimeoutException;
 
 import static ai.vespa.metricsproxy.TestUtil.getFileContents;
+import static ai.vespa.metricsproxy.http.application.ApplicationMetricsRetriever.MAX_THREADS;
+import static ai.vespa.metricsproxy.http.application.ApplicationMetricsRetriever.MAX_TIMEOUT;
+import static ai.vespa.metricsproxy.http.application.ApplicationMetricsRetriever.MIN_TIMEOUT;
+import static ai.vespa.metricsproxy.http.application.ApplicationMetricsRetriever.timeout;
 import static com.github.tomakehurst.wiremock.client.WireMock.aResponse;
 import static com.github.tomakehurst.wiremock.client.WireMock.get;
 import static com.github.tomakehurst.wiremock.client.WireMock.urlPathEqualTo;
@@ -146,6 +150,17 @@ public class ApplicationMetricsRetrieverTest {
         verifyRetrievingMetricsFromSingleNode(config, node);
     }
 
+    @Test
+    public void test_timeout_calculation() {
+        assertEquals(MIN_TIMEOUT, timeout(1, 1));
+        assertEquals(MIN_TIMEOUT, timeout(MAX_THREADS, MAX_THREADS));
+
+        // These values must be updated if the calculation in the timeout method itself is changed.
+        assertEquals(Duration.ofSeconds(100), timeout(100, MAX_THREADS));
+        assertEquals(Duration.ofSeconds(200), timeout(200, MAX_THREADS));
+        assertEquals(MAX_TIMEOUT, timeout(240, MAX_THREADS));
+    }
+
     private MetricsNodesConfig nodesConfig(String... paths) {
         var nodes = Arrays.stream(paths)
                 .map(this::nodeConfig)