Fix JENKINS-8196

Persist a user-requested offline status for a slave across restarts. Previously slaves would come back online if the server was restarted while they were marked 'temporarily offline'. I believe this also addresses JENKINS-7934 and JENKINS-7805 which look like duplicates.
jenkinsci · Feb 4, 2011 · a88920a · a88920a
1 parent d990fbb
commit a88920a
Show file tree

Hide file tree

Showing 2 changed files with 53 additions and 0 deletions.
diff --git a/core/src/main/java/hudson/model/Computer.java b/core/src/main/java/hudson/model/Computer.java
@@ -509,6 +509,7 @@ public void setTemporarilyOffline(boolean temporarilyOffline) {
     public void setTemporarilyOffline(boolean temporarilyOffline, OfflineCause cause) {
         offlineCause = temporarilyOffline ? cause : null;
         this.temporarilyOffline = temporarilyOffline;
+        getNode().setTemporaryOfflineCause(offlineCause);
         Hudson.getInstance().getQueue().scheduleMaintenance();
     }
 
@@ -563,6 +564,16 @@ protected void setNode(Node node) {
             this.nodeName = null;
 
         setNumExecutors(node.getNumExecutors());
+        if (this.temporarilyOffline) {
+            // When we get a new node, push our current temp offline
+            // status to it (as the status is not carried across
+            // configuration changes that recreate the node).
+            // Since this is also called the very first time this
+            // Computer is created, avoid pushing an empty status
+            // as that could overwrite any status that the Node
+            // brought along from its persisted config data.
+            node.setTemporaryOfflineCause(this.offlineCause);
+        }
     }
 
     /**

diff --git a/core/src/main/java/hudson/model/Node.java b/core/src/main/java/hudson/model/Node.java
@@ -25,6 +25,7 @@
 package hudson.model;
 
 import com.infradna.tool.bridge_method_injector.WithBridgeMethods;
+import hudson.Extension;
 import hudson.ExtensionPoint;
 import hudson.FilePath;
 import hudson.FileSystemProvisioner;
@@ -37,9 +38,11 @@
 import hudson.security.ACL;
 import hudson.security.AccessControlled;
 import hudson.security.Permission;
+import hudson.slaves.ComputerListener;
 import hudson.slaves.NodeDescriptor;
 import hudson.slaves.NodeProperty;
 import hudson.slaves.NodePropertyDescriptor;
+import hudson.slaves.OfflineCause;
 import hudson.util.ClockDifference;
 import hudson.util.DescribableList;
 import hudson.util.EnumConverter;
@@ -51,6 +54,7 @@
 import java.util.HashSet;
 import java.util.Set;
 import java.util.List;
+import java.util.logging.Logger;
 
 import org.kohsuke.stapler.Stapler;
 import org.kohsuke.stapler.export.ExportedBean;
@@ -68,6 +72,9 @@
  */
 @ExportedBean
 public abstract class Node extends AbstractModelObject implements Describable<Node>, ExtensionPoint, AccessControlled {
+
+    private static final Logger LOGGER = Logger.getLogger(Node.class.getName());
+
     /**
      * Newly copied slaves get this flag set, so that Hudson doesn't try to start this node until its configuration
      * is saved once.
@@ -165,6 +172,41 @@ public final VirtualChannel getChannel() {
      */
     protected abstract Computer createComputer();
 
+    /**
+     * Let Nodes be aware of the lifecycle of their own {@link Computer}.
+     */
+    @Extension
+    public static class InternalComputerListener extends ComputerListener {
+        @Override
+        public void onOnline(Computer c, TaskListener listener) {
+            Node node = c.getNode();
+
+            // At startup, we need to restore any previously in-effect temp offline cause.
+            // We wait until the computer is started rather than getting the data to it sooner
+            // so that the normal computer start up processing works as expected.
+            if (node.temporaryOfflineCause != null && node.temporaryOfflineCause != c.getOfflineCause()) {
+                c.setTemporarilyOffline(true, node.temporaryOfflineCause);
+            }
+        }
+    }
+
+    private OfflineCause temporaryOfflineCause;
+
+    /**
+     * Enable a {@link Computer} to inform its node when it is taken
+     * temporarily offline.
+     */
+    void setTemporaryOfflineCause(OfflineCause cause) {
+        try {
+            if (temporaryOfflineCause != cause) {
+                temporaryOfflineCause = cause;
+                Hudson.getInstance().save(); // Gotta be a better way to do this
+            }
+        } catch (java.io.IOException e) {
+            LOGGER.warning("Unable to complete save, temporary offline status will not be persisted: " + e.getMessage());
+        }
+    }
+
     /**
      * Return the possibly empty tag cloud for the labels of this node.
      */