Prevent premature removal of agent because of race condition in Docke…

…rSwarmAgentRetentionStrategy (#115) Co-authored-by: ue56923 <[email protected]> Co-authored-by: Roman <[email protected]>
jenkinsci · Jan 4, 2023 · d3f18e9 · d3f18e9
1 parent 2b768b1
commit d3f18e9
Show file tree

Hide file tree

Showing 2 changed files with 25 additions and 1 deletion.
diff --git a/src/main/java/org/jenkinsci/plugins/docker/swarm/DockerSwarmAgentRetentionStrategy.java b/src/main/java/org/jenkinsci/plugins/docker/swarm/DockerSwarmAgentRetentionStrategy.java
@@ -44,8 +44,9 @@ public long getTimeout() {
     public long check(@Nonnull DockerSwarmComputer c) {
         if (c.isIdle() && c.isOnline()) {
             final long connectTime = System.currentTimeMillis() - c.getConnectTime();
+            final long onlineTime = System.currentTimeMillis() - c.getOnlineTime();
             final long idleTime = System.currentTimeMillis() - c.getIdleStartMilliseconds();
-            final boolean isTimeout = connectTime > timeout && idleTime > timeout;
+            final boolean isTimeout = connectTime > timeout && onlineTime > timeout && idleTime > timeout;
             if (isTimeout && (!isTaskAccepted || isTaskCompleted ) && !Jenkins.getInstance().isQuietingDown()) {
                 LOGGER.log(Level.INFO, "Disconnecting due to idle {0}", c.getName());
                 done(c);

diff --git a/src/main/java/org/jenkinsci/plugins/docker/swarm/DockerSwarmComputer.java b/src/main/java/org/jenkinsci/plugins/docker/swarm/DockerSwarmComputer.java
@@ -1,19 +1,29 @@
 
 package org.jenkinsci.plugins.docker.swarm;
 
+import java.io.IOException;
+import java.io.OutputStream;
+
 import java.util.HashMap;
 import java.util.Map;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.Future;
+import java.util.logging.Level;
+import java.util.logging.Logger;
 
 import com.google.common.collect.Iterables;
 
 import hudson.model.Executor;
 import hudson.model.Queue;
+import hudson.remoting.Channel;
+import hudson.remoting.Channel.Listener;
 import hudson.slaves.AbstractCloudComputer;
 import hudson.slaves.OfflineCause;
 
 public class DockerSwarmComputer extends AbstractCloudComputer<DockerSwarmAgent> {
+    private static final Logger LOGGER = Logger.getLogger(DockerSwarmComputer.class.getName());
+
+    private long onlineTime = 0L;
 
     public DockerSwarmComputer(final DockerSwarmAgent dockerSwarmAgent) {
         super(dockerSwarmAgent);
@@ -55,4 +65,17 @@ public Future<?> disconnect(OfflineCause cause) {
     public String getVolumeName() {
         return getName().split("-")[1];
     }
+
+    public final long getOnlineTime() {
+        return onlineTime;
+    }
+
+    @Override
+    public void setChannel(Channel channel, OutputStream launchLog, Listener listener) throws IOException, InterruptedException {
+        this.onlineTime = System.currentTimeMillis();
+
+        super.setChannel(channel, launchLog, listener);
+
+        LOGGER.log(Level.INFO, "Agent {0} got online", getName());
+    }
 }