Ensure that default healthcheck duration is less than node timeout du…

…ration If the healthcheck does not pass before the node timeout period is passed, the node is marked as being down. The default timings ensured this happened, which is distinctly suboptimal.
regularfry · Jul 6, 2021 · 0795e78 · 0795e78
1 parent 9b1f2d7
commit 0795e78
Show file tree

Hide file tree

Showing 2 changed files with 13 additions and 3 deletions.
diff --git a/java/server/src/org/openqa/selenium/grid/distributor/config/DistributorOptions.java b/java/server/src/org/openqa/selenium/grid/distributor/config/DistributorOptions.java
@@ -31,7 +31,7 @@
 
 public class DistributorOptions {
 
-  public static final int DEFAULT_HEALTHCHECK_INTERVAL = 300;
+  public static final int DEFAULT_HEALTHCHECK_INTERVAL = 120;
   public static final String DISTRIBUTOR_SECTION = "distributor";
   static final String DEFAULT_DISTRIBUTOR_IMPLEMENTATION =
     "org.openqa.selenium.grid.distributor.local.LocalDistributor";

diff --git a/java/server/src/org/openqa/selenium/grid/distributor/local/LocalDistributor.java b/java/server/src/org/openqa/selenium/grid/distributor/local/LocalDistributor.java
@@ -58,6 +58,7 @@
 import org.openqa.selenium.grid.sessionmap.config.SessionMapOptions;
 import org.openqa.selenium.grid.sessionqueue.NewSessionQueue;
 import org.openqa.selenium.grid.sessionqueue.config.NewSessionQueueOptions;
+import org.openqa.selenium.internal.Debug;
 import org.openqa.selenium.internal.Either;
 import org.openqa.selenium.internal.Require;
 import org.openqa.selenium.remote.SessionId;
@@ -249,8 +250,6 @@ private void register(NodeStatus status) {
   public LocalDistributor add(Node node) {
     Require.nonNull("Node", node);
 
-    LOG.info(String.format("Added node %s at %s.", node.getId(), node.getUri()));
-
     nodes.put(node.getId(), node);
     model.add(node.getStatus());
 
@@ -259,6 +258,12 @@ public LocalDistributor add(Node node) {
     allChecks.put(node.getId(), runnableHealthCheck);
     hostChecker.submit(runnableHealthCheck, healthcheckInterval, Duration.ofSeconds(30));
 
+    LOG.info(String.format(
+      "Added node %s at %s. Health check every %ss",
+      node.getId(),
+      node.getUri(),
+      healthcheckInterval.toMillis() / 1000));
+
     bus.fire(new NodeAddedEvent(node.getId()));
 
     return this;
@@ -268,6 +273,8 @@ private Runnable asRunnableHealthCheck(Node node) {
     HealthCheck healthCheck = node.getHealthCheck();
     NodeId id = node.getId();
     return () -> {
+      LOG.log(getDebugLogLevel(), "Running health check for " + node.getId());
+
       HealthCheck.Result result;
       try {
         result = healthCheck.check();
@@ -279,6 +286,9 @@ private Runnable asRunnableHealthCheck(Node node) {
       Lock writeLock = lock.writeLock();
       writeLock.lock();
       try {
+        LOG.log(
+          getDebugLogLevel(),
+          String.format("Health check result for %s was %s", node.getId(), result.getAvailability()));
         model.setAvailability(id, result.getAvailability());
       } finally {
         writeLock.unlock();