From c6f31bfb56b866f05c7def172af96753136a35cf Mon Sep 17 00:00:00 2001
From: res0nance <raihaanhimself@gmail.com>
Date: Sat, 2 Sep 2023 22:15:38 +0800
Subject: [PATCH] JENKINS-71554: Reconnect on "RequestExpired" exception

---
 .../java/hudson/plugins/ec2/EC2Cloud.java     | 43 +++++++++++--------
 1 file changed, 25 insertions(+), 18 deletions(-)

diff --git a/src/main/java/hudson/plugins/ec2/EC2Cloud.java b/src/main/java/hudson/plugins/ec2/EC2Cloud.java
index f20644f0e..95a04b4a6 100644
--- a/src/main/java/hudson/plugins/ec2/EC2Cloud.java
+++ b/src/main/java/hudson/plugins/ec2/EC2Cloud.java
@@ -19,6 +19,7 @@
 package hudson.plugins.ec2;
 
 import com.amazonaws.AmazonClientException;
+import com.amazonaws.AmazonServiceException;
 import com.amazonaws.ClientConfiguration;
 import com.amazonaws.auth.AWSCredentials;
 import com.amazonaws.auth.AWSCredentialsProvider;
@@ -694,7 +695,7 @@ private int getPossibleNewSlavesCount(SlaveTemplate template) throws AmazonClien
      * Obtains a agent whose AMI matches the AMI of the given template, and that also has requiredLabel (if requiredLabel is non-null)
      * forceCreateNew specifies that the creation of a new agent is required. Otherwise, an existing matching agent may be re-used
      */
-    private List<EC2AbstractSlave> getNewOrExistingAvailableSlave(SlaveTemplate t, int number, boolean forceCreateNew) {
+    private List<EC2AbstractSlave> getNewOrExistingAvailableSlave(SlaveTemplate t, int number, boolean forceCreateNew) throws IOException {
         try {
             slaveCountingLock.lock();
             int possibleSlavesCount = getPossibleNewSlavesCount(t);
@@ -703,24 +704,19 @@ private List<EC2AbstractSlave> getNewOrExistingAvailableSlave(SlaveTemplate t, i
                 return null;
             }
 
-            try {
-                EnumSet<SlaveTemplate.ProvisionOptions> provisionOptions;
-                if (forceCreateNew)
-                    provisionOptions = EnumSet.of(SlaveTemplate.ProvisionOptions.FORCE_CREATE);
-                else
-                    provisionOptions = EnumSet.of(SlaveTemplate.ProvisionOptions.ALLOW_CREATE);
-
-                if (number > possibleSlavesCount) {
-                    LOGGER.log(Level.INFO, String.format("%d nodes were requested for the template %s, " +
-                            "but because of instance cap only %d can be provisioned", number, t, possibleSlavesCount));
-                    number = possibleSlavesCount;
-                }
+            EnumSet<SlaveTemplate.ProvisionOptions> provisionOptions;
+            if (forceCreateNew)
+                provisionOptions = EnumSet.of(SlaveTemplate.ProvisionOptions.FORCE_CREATE);
+            else
+                provisionOptions = EnumSet.of(SlaveTemplate.ProvisionOptions.ALLOW_CREATE);
 
-                return t.provision(number, provisionOptions);
-            } catch (IOException e) {
-                LOGGER.log(Level.WARNING, t + ". Exception during provisioning", e);
-                return null;
+            if (number > possibleSlavesCount) {
+                LOGGER.log(Level.INFO, String.format("%d nodes were requested for the template %s, " +
+                        "but because of instance cap only %d can be provisioned", number, t, possibleSlavesCount));
+                number = possibleSlavesCount;
             }
+
+            return t.provision(number, provisionOptions);
         } finally { slaveCountingLock.unlock(); }
     }
 
@@ -761,7 +757,18 @@ public Collection<PlannedNode> provision(final Label label, int excessWorkload)
 
                 LOGGER.log(Level.INFO, "{0}. Attempting provision finished, excess workload: " + excessWorkload, t);
                 if (excessWorkload == 0) break;
-            } catch (AmazonClientException e) {
+            } catch (AmazonServiceException e) {
+                LOGGER.log(Level.WARNING, t + ". Exception during provisioning", e);
+                if (e.getErrorCode().equals("RequestExpired")) {
+                    // JENKINS-71554: A RequestExpired error can indicate that credentials have expired so reconnect
+                    LOGGER.log(Level.INFO, "[JENKINS-71554] Reconnecting to EC2 due to RequestExpired error");
+                    try {
+                        reconnectToEc2();
+                    } catch (IOException e2) {
+                        LOGGER.log(Level.WARNING, "Failed to reconnect ec2", e2);
+                    }
+                }
+            } catch (AmazonClientException | IOException e) {
                 LOGGER.log(Level.WARNING, t + ". Exception during provisioning", e);
             }
         }