[DSIP-17122][SERVER] Kubernetes Connection Pool Implementation #17510

Cat-Drink · 2025-09-15T11:20:42Z

Purpose of the pull request

This pull request introduces support for the Kubernetes client connection pool, optimizing the interaction performance and resource management between Apache DolphinScheduler and Kubernetes clusters. By maintaining a reuse mechanism for client connections, this feature significantly reduces the overhead caused by frequent creation and destruction of Kubernetes clients, thereby improving the overall efficiency and stability of the system.

Brief change log

Core Backend Implementation: Added the KubernetesClientPool class, which implements a singleton-pattern connection pool manager to maintain an independent connection pool for each Kubernetes cluster.
Connection Pool Configuration: Implemented the PoolConfig inner class, which supports configuration items such as minimum idle connections, maximum idle connections, maximum connections, maximum waiting time, and idle timeout. These configurations can be customized through properties in TaskConstants.
Cluster Connection Pool: Implemented the ClusterClientPool inner class, which is responsible for managing client connections of specific clusters, including the creation, acquisition, return, and closure of connections.
Idle Connection Cleanup: Added a periodic cleanup thread that executes the cleanupIdleClients method every 30 seconds to clear timed-out idle connections, effectively saving resources.
Cluster Identification: Implemented a cluster identification generation mechanism based on kubeconfig, using the SHA-256 algorithm to generate a unique identifier for each cluster.
Client Validation: Added the isClientValid method to ensure that the returned client connection remains valid.
Integration into Utility Classes: Updated utility classes such as K8sUtils and KubernetesApplicationManager to enable them to acquire and return clients using the connection pool.

Verify this pull request

This change includes comprehensive test cases and can be verified through the following methods:

Automated Testing

Basic Functionality Test: KubernetesClientPoolTest.testKubernetesClientPoolBasicFunction verifies the basic operations of the connection pool (acquiring cluster ID, obtaining client, returning client, and closing the connection pool).
Configuration Test: KubernetesClientPoolTest.testKubernetesClientPoolConfig verifies the correctness of connection pool configurations.
Cluster ID Generation Test: KubernetesClientPoolTest.testClusterIdGeneration verifies the function of generating a unique cluster identifier based on kubeconfig.
Singleton Pattern Test: KubernetesClientPoolTest.testSingletonPattern verifies the singleton pattern implementation of the connection pool.
Concurrent Access Test: Verifies the thread safety and stability of the connection pool in a multi-threaded environment.
Timeout Handling Test: KubernetesClientPoolTest.testClusterClientPoolBorrowObjectTimeout verifies the handling mechanism for client acquisition timeouts.
Idle Connection Cleanup Test: KubernetesClientPoolTest.testCleanupIdleClients and KubernetesClientPoolTest.testClusterClientPoolCleanupIdle verify the periodic cleanup function of idle connections.

Manual Verification Steps

Ensure that appropriate connection pool parameters (such as k8s.client.pool.max.size) are set in the configuration file.
Submit multiple Kubernetes tasks to different clusters.
Check the logs to verify whether the connection pool correctly creates and manages client connections.
Monitor system resource usage to verify whether the connection pool effectively reduces the overhead caused by connection creation and destruction.

Pull Request Notice

If your pull request contains incompatible change, you should also add it to docs/docs/en/guide/upgrade/incompatible.md

…ation

… and adapted to special handling in Watch scenarios

…connection pool interface of K8sUtils through the configYaml parameter

boring-cyborg · 2025-09-15T11:20:44Z

Thanks for opening this pull request! Please check out our contributing guidelines. (https://github.com/apache/dolphinscheduler/blob/dev/docs/docs/en/contribute/join/pull-request.md)

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

SbloodyS

This pr's title and DSIP-88 are no related. Please modify the correct title and with full design detail and description. @Cat-Drink

Cat-Drink · 2025-10-26T05:05:54Z

PTAL‌

SbloodyS · 2025-10-28T02:09:07Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+        String kubeConfig = k8sTaskExecutionContext.getConfigYaml();
+        int hashCode = kubeConfig.hashCode();
+        int nonNegativeHash = hashCode & 0x7FFFFFFF;
+        return "k8s-cluster-" + nonNegativeHash;


It's better to use constants here.

SbloodyS · 2025-10-28T02:09:34Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+            byte[] hashBytes = digest.digest(kubeConfig.getBytes(StandardCharsets.UTF_8));
+
+            String base64Hash = Base64.getUrlEncoder().encodeToString(hashBytes);
+            return base64Hash.replace("=", "");


It's better to use constants here.

SbloodyS · 2025-10-28T02:10:05Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+            while (true) {
+                try {
+                    // every 30s
+                    Thread.sleep(30000);


It's better to use constants here.

SbloodyS · 2025-10-28T02:10:11Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+        Thread cleanupThread = new Thread(() -> {
+            while (true) {
+                try {
+                    // every 30s


Suggested change

// every 30s

SbloodyS · 2025-10-28T02:10:45Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+        public int getMaxSize() {
+            return maxSize;
+        }
+        public int getMinIdle() {
+            return minIdle;
+        }
+        public int getMaxIdle() {
+            return maxIdle;
+        }
+        public long getMaxWaitMs() {
+            return maxWaitMs;
+        }
+        public long getIdleTimeoutMs() {
+            return idleTimeoutMs;
+        }


Using @Getter here.

SbloodyS · 2025-10-28T02:12:57Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/impl/K8sTaskExecutor.java

        } catch (Exception e) {
            log.error("job failed in k8s: {}", e.getMessage(), e);
            taskResponse.setExitStatusCode(EXIT_CODE_FAILURE);
+            log.error(e.getMessage());


Suggested change

log.error(e.getMessage());

SbloodyS · 2025-10-28T02:13:06Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/impl/K8sTaskExecutor.java

                }
            }
        } catch (Exception e) {
+            // if Exception happen


Suggested change

// if Exception happen

SbloodyS · 2025-10-28T02:14:00Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/impl/K8sTaskExecutor.java

-            throw e;
-        } finally {
-            ProcessUtils.removeK8sClientCache(taskRequest.getTaskAppId());
+            log.error(e.getMessage());


Suggested change

log.error(e.getMessage());

log.error("k8s task executor running error: ", e);

SbloodyS · 2025-10-28T02:14:41Z

...sk-api/src/test/java/org/apache/dolphinscheduler/plugin/task/api/utils/ProcessUtilsTest.java

+    /**
+     * 测试K8s客户端是否被正确归还到连接池
+     */


Please avoid using chinese.

SbloodyS · 2025-10-28T02:15:23Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+        this.poolConfig = new PoolConfig(
+                PropertyUtils.getInt(TaskConstants.K8S_CLIENT_POOL_MAX_SIZE, 10),
+                PropertyUtils.getInt(TaskConstants.K8S_CLIENT_POOL_MIN_IDLE, 2),
+                PropertyUtils.getInt(TaskConstants.K8S_CLIENT_POOL_MAX_IDLE, 5),
+                PropertyUtils.getInt(TaskConstants.K8S_CLIENT_POOL_MAX_WAIT_MS, 30000),
+                PropertyUtils.getInt(TaskConstants.K8S_CLIENT_POOL_IDLE_TIMEOUT_MS, 600000));


These params should add to docs and guide user how to use it.

SbloodyS · 2025-10-28T02:16:21Z

This pr's title and DSIP-88 are no related.

BTW, this is still unaddressed.

ruanwenjun · 2025-10-30T05:21:20Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+                    .inContainer(kubernetesApplicationManagerContext.getContainerName())
+                    .watchLog();
+        } finally {
+            log.debug("Log watch client is not returned immediately, will be managed by caller after watch completes");


This log seems meaningless.

Copilot

Pull Request Overview

This PR introduces a Kubernetes client connection pool (KubernetesClientPool) to manage K8s client connections efficiently across multiple clusters. The changes replace the previous approach of creating individual clients and caching them per task with a robust pooling mechanism that supports connection reuse, lifecycle management, and automatic cleanup of idle connections.

Key Changes

Implements a new KubernetesClientPool singleton with per-cluster connection pooling
Refactors K8sUtils to use the connection pool instead of instance-level clients
Updates KubernetesApplicationManager to retrieve clients from the pool and return them after use
Removes the removeK8sClientCache method from ProcessUtils as caching is now handled by the pool
Adds comprehensive test coverage for the new pooling functionality

Reviewed Changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
KubernetesClientPool.java	New connection pool implementation with cluster-based pooling, idle cleanup, and configurable limits
K8sUtils.java	Refactored to use connection pool; all methods now accept `configYaml` parameter and properly return clients
KubernetesApplicationManager.java	Updated to use connection pool instead of per-task client caching
K8sTaskExecutor.java	Modified to pass `configYaml` to K8sUtils methods and removed client build calls
ProcessUtils.java	Removed `removeK8sClientCache` method
TaskConstants.java	Added configuration constants for pool sizing and timeouts
KubernetesClientPoolTest.java	Comprehensive test suite for connection pool functionality
K8sUtilsTest.java	New tests for refactored K8sUtils methods
KubernetesApplicationManagerTest.java	Tests for application manager with connection pool
K8sTaskExecutorTest.java	Updated tests with proper mocking for connection pool usage
configuration.md (en/zh)	Documentation for new connection pool configuration parameters

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-10-30T05:25:58Z

...r-task-api/src/test/java/org/apache/dolphinscheduler/plugin/task/api/utils/K8sUtilsTest.java

+    @BeforeEach
+    public void setUp() {
+        k8sUtils = new K8sUtils();
+        mockedKubernetesClientPool = Mockito.mockStatic(KubernetesClientPool.class); // 拦截所有使用静态方法的请求


Chinese comment should be removed or translated to English. The comment '拦截所有使用静态方法的请求' (intercept all requests using static methods) is inconsistent with the rest of the codebase which uses English comments.

Suggested change

mockedKubernetesClientPool = Mockito.mockStatic(KubernetesClientPool.class); // 拦截所有使用静态方法的请求

mockedKubernetesClientPool = Mockito.mockStatic(KubernetesClientPool.class); // Intercept all requests using static methods

Copilot · 2025-10-30T05:25:59Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/impl/K8sTaskExecutor.java


    @Override
-    public TaskResponse run(String k8sParameterStr) throws Exception {
+    public TaskResponse run(String k8sParameterStr) {


The method signature changed from throws Exception to no exception declaration, but the method body still contains try-catch blocks that catch Exception. This is a breaking API change that could affect callers expecting to handle exceptions. The removed exception should either be kept in the signature or properly handled/logged in all code paths.

Suggested change

public TaskResponse run(String k8sParameterStr) {

public TaskResponse run(String k8sParameterStr) throws Exception {

Copilot · 2025-10-30T05:25:59Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+                KubernetesClient client = null;
+                try {
+                    client = getClient(kubernetesApplicationManagerContext);
+                    // Retrieve watchList again, as the previous instance of tes client connection pool may have expired


Typo in comment: 'tes client' should be 'the client'.

Suggested change

// Retrieve watchList again, as the previous instance of tes client connection pool may have expired

// Retrieve watchList again, as the previous instance of the client connection pool may have expired

Copilot · 2025-10-30T05:25:59Z

...-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPool.java

+            return base64Hash.replace(BASE64_PADDING_CHARACTER, EMPTY_STRING);
+        } catch (Exception e) {
+            log.error("Failed to generate cluster ID", e);
+            return Integer.toString(kubeConfig.hashCode());


The fallback cluster ID generation using hashCode() can produce negative values, which could cause issues. Although line 170 in KubernetesApplicationManager uses hashCode & 0x7FFFFFFF to ensure non-negative values, this fallback in KubernetesClientPool doesn't apply the same mask, creating inconsistency in cluster ID generation.

Suggested change

return Integer.toString(kubeConfig.hashCode());

return Integer.toString(kubeConfig.hashCode() & 0x7FFFFFFF);

Copilot · 2025-10-30T05:26:00Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+        String kubeConfig = k8sTaskExecutionContext.getConfigYaml();
+        int hashCode = kubeConfig.hashCode();
+        int nonNegativeHash = hashCode & 0x7FFFFFFF;
+        return K8S_CLUSTER_PREFIX + nonNegativeHash;


The cluster ID generation logic is duplicated between KubernetesApplicationManager.getClusterId() and KubernetesClientPool.getClusterId(). These methods use different algorithms (SHA-256 in the pool vs hashCode in the manager), which will produce different cluster IDs for the same kubeConfig. This inconsistency will prevent proper client pooling and cause connection leaks.

Copilot · 2025-10-30T05:26:00Z

...duler-task-api/src/main/java/org/apache/dolphinscheduler/plugin/task/api/utils/K8sUtils.java

+                log.debug(
+                        "createBatchJobWatcher does not return client immediately, caller should manage client lifecycle");


The comment indicates the client is not returned immediately and the caller should manage the lifecycle, but the client is obtained in the try block (line 97) and logged in the finally block (line 106-109) without actually being returned to the pool. This will cause a client connection leak as the client remains borrowed but is never returned.

Copilot · 2025-10-30T05:26:00Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+            return client.pods().inNamespace(pod.getMetadata().getNamespace())
+                    .withName(pod.getMetadata().getName())
+                    .inContainer(kubernetesApplicationManagerContext.getContainerName())
+                    .watchLog();
+        } finally {
+            log.debug("Log watch client is not returned immediately, will be managed by caller after watch completes");
+        }
    }



Similar to the issue in K8sUtils.createBatchJobWatcher, the client obtained in getPodLogWatcher (line 263) is not returned to the pool in the finally block. The debug message suggests it's intentional, but this creates a connection leak as there's no guarantee the caller will properly return the client after the watch completes.

Suggested change

return client.pods().inNamespace(pod.getMetadata().getNamespace())

.withName(pod.getMetadata().getName())

.inContainer(kubernetesApplicationManagerContext.getContainerName())

.watchLog();

} finally {

log.debug("Log watch client is not returned immediately, will be managed by caller after watch completes");

}

}

LogWatch logWatch = client.pods().inNamespace(pod.getMetadata().getNamespace())

.withName(pod.getMetadata().getName())

.inContainer(kubernetesApplicationManagerContext.getContainerName())

.watchLog();

return new ClientReturningLogWatch(logWatch, client, kubernetesApplicationManagerContext);

} catch (Exception e) {

if (client != null) {

KubernetesClientPool.returnClient(kubernetesApplicationManagerContext, client);

}

throw e;

}

}

/**

* Wrapper for LogWatch that returns the client to the pool when closed.

*/

private static class ClientReturningLogWatch implements LogWatch {

private final LogWatch delegate;

private final KubernetesClient client;

private final KubernetesApplicationManagerContext context;

private boolean closed = false;

public ClientReturningLogWatch(LogWatch delegate, KubernetesClient client, KubernetesApplicationManagerContext context) {

this.delegate = delegate;

this.client = client;

this.context = context;

}

@Override

public void close() {

if (!closed) {

try {

delegate.close();

} finally {

KubernetesClientPool.returnClient(context, client);

closed = true;

}

}

}

@Override

public java.io.InputStream getOutput() {

return delegate.getOutput();

}

@Override

public java.io.InputStream getError() {

return delegate.getError();

}

}

Copilot · 2025-10-30T05:26:00Z

.../src/test/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPoolTest.java

+            KubernetesClient client1 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);
+            KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);


Variable 'KubernetesClient client1' is never read.

Suggested change

KubernetesClient client1 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);

KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);

borrowObjectMethod.invoke(clusterClientPool);

borrowObjectMethod.invoke(clusterClientPool);

Copilot · 2025-10-30T05:26:01Z

.../src/test/java/org/apache/dolphinscheduler/plugin/task/api/k8s/KubernetesClientPoolTest.java

+
+            // Get all available clients (maxSize = 2)
+            KubernetesClient client1 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);
+            KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);


Variable 'KubernetesClient client2' is never read.

Suggested change

KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);

borrowObjectMethod.invoke(clusterClientPool);

wangxj3 · 2025-09-16T06:11:10Z

...c/main/java/org/apache/dolphinscheduler/plugin/task/api/am/KubernetesApplicationManager.java

+                String labelValue = kubernetesApplicationManagerContext.getLabelValue();
+
+                // Build the latest watchList again
+                FilterWatchListDeletable<Pod, PodList, PodResource> newWatchList =


this is a new newWatchList?

yes, it is a newWatchList

sonarqubecloud · 2025-10-30T08:16:06Z

Quality Gate passed

Issues
74 New issues
0 Accepted issues

Measures
0 Security Hotspots
75.4% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

Cat-Drink added 8 commits September 15, 2025 10:20

KubernetesClientPool.java: Added a complete connection pool implement…

8c254ad

…ation

K8sUtils.java: Updated to k8s connection pool client management logic…

9dbc37d

… and adapted to special handling in Watch scenarios

K8sTaskExecutor. Java: Adjust all K8s operation methods and call the …

59c20bf

…connection pool interface of K8sUtils through the configYaml parameter

TaskConstants.java:Kubernetes client pool configuration

4dc7694

KubernetesApplicationManager.java:fix k8s pool manager

841880b

fix:spotless:apply

31ac1f6

K8sTaskExecutorTest.java:add k8s pool unitTest

3adb349

unitTest spotless:apply

63d3066

Cat-Drink requested review from Gallardot, SbloodyS, caishunfeng and songjianet as code owners September 15, 2025 11:20

github-actions bot added UI ui and front end related backend test labels Sep 15, 2025

github-actions bot assigned Cat-Drink Sep 15, 2025

Cat-Drink changed the title ~~K8s Connections Pool~~ [Feature-3328][Task] Kubernetes Connection Pool Implementation Sep 15, 2025

wangxj3 changed the title ~~[Feature-3328][Task] Kubernetes Connection Pool Implementation~~ [DSIP-88][Task] Kubernetes Connection Pool Implementation Sep 16, 2025

Cat-Drink added 4 commits September 16, 2025 12:29

fix:K8sTaskExecutorTest.java

42c432d

fix:ProcessUtilsTest.java

c579ea6

ProcessUtilsTest spotless:apply

39a2035

reuse pnpm-lock.yaml

5742fc2

github-advanced-security bot found potential problems Sep 16, 2025

View reviewed changes

fix:Code scanning

7d4ce20

SbloodyS requested changes Sep 17, 2025

View reviewed changes

Cat-Drink added 4 commits September 17, 2025 13:08

Optimize K8sUtils,KubernetesClientPool

bf686d9

add:K8sUtilsTest.java

a502799

use spotless:apply

4b4fab2

remove chinese

d86932f

Merge branch 'dev' into dev

29f0ffc

Merge branch 'dev' into dev

5872bad

davidzollo added the improvement make more easy to user or prompt friendly label Oct 28, 2025

SbloodyS requested changes Oct 28, 2025

View reviewed changes

wangxj3 changed the title ~~[Improvement-88][SERVER] Kubernetes Connection Pool Implementation~~ [DSIP-88][SERVER] Kubernetes Connection Pool Implementation Oct 28, 2025

Cat-Drink added 4 commits October 29, 2025 11:42

docs: about k8s connection pool settings

bc04e78

fix: spotless:apply

d967d3f

fix: related review

cc7c033

Merge remote-tracking branch 'origin/dev' into dev

96c9bd7

github-actions bot added the document label Oct 29, 2025

Cat-Drink changed the title ~~[DSIP-88][SERVER] Kubernetes Connection Pool Implementation~~ [DSIP-17122][SERVER] Kubernetes Connection Pool Implementation Oct 29, 2025

nielifeng requested a review from Copilot October 30, 2025 05:16

ruanwenjun reviewed Oct 30, 2025

View reviewed changes

Copilot AI reviewed Oct 30, 2025

View reviewed changes

wangxj3 and others added 3 commits October 30, 2025 14:08

Merge branch 'dev' into dev

5050d2a

fix: reback throw exception

9452780

fix: getClusterId duplicated and add returnClient

99ae23e

wangxj3 approved these changes Oct 30, 2025

View reviewed changes

wangxj3 previously approved these changes Oct 30, 2025

View reviewed changes

Cat-Drink added 2 commits October 30, 2025 15:02

fix: some issue about Test

0169adc

Merge remote-tracking branch 'origin/dev' into dev

26ad45d

Cat-Drink dismissed wangxj3’s stale review via 26ad45d October 30, 2025 07:10

spotless:apply

87050b8

Cat-Drink requested review from SbloodyS, ruanwenjun and wangxj3 October 31, 2025 02:23

SbloodyS closed this Nov 14, 2025

	log.error(e.getMessage());
	log.error("k8s task executor running error: ", e);

	mockedKubernetesClientPool = Mockito.mockStatic(KubernetesClientPool.class); // 拦截所有使用静态方法的请求
	mockedKubernetesClientPool = Mockito.mockStatic(KubernetesClientPool.class); // Intercept all requests using static methods

	public TaskResponse run(String k8sParameterStr) {
	public TaskResponse run(String k8sParameterStr) throws Exception {

	// Retrieve watchList again, as the previous instance of tes client connection pool may have expired
	// Retrieve watchList again, as the previous instance of the client connection pool may have expired

	return Integer.toString(kubeConfig.hashCode());
	return Integer.toString(kubeConfig.hashCode() & 0x7FFFFFFF);

		log.debug(
		"createBatchJobWatcher does not return client immediately, caller should manage client lifecycle");

-            return client.pods().inNamespace(pod.getMetadata().getNamespace())
-                    .withName(pod.getMetadata().getName())
-                    .inContainer(kubernetesApplicationManagerContext.getContainerName())
-                    .watchLog();
-        } finally {
-            log.debug("Log watch client is not returned immediately, will be managed by caller after watch completes");
-        }
-    }
+            LogWatch logWatch = client.pods().inNamespace(pod.getMetadata().getNamespace())
+                    .withName(pod.getMetadata().getName())
+                    .inContainer(kubernetesApplicationManagerContext.getContainerName())
+                    .watchLog();
+            return new ClientReturningLogWatch(logWatch, client, kubernetesApplicationManagerContext);
+        } catch (Exception e) {
+            if (client != null) {
+                KubernetesClientPool.returnClient(kubernetesApplicationManagerContext, client);
+            }
+            throw e;
+        }
+    }
+    /**
+     * Wrapper for LogWatch that returns the client to the pool when closed.
+     */
+    private static class ClientReturningLogWatch implements LogWatch {
+        private final LogWatch delegate;
+        private final KubernetesClient client;
+        private final KubernetesApplicationManagerContext context;
+        private boolean closed = false;
+        public ClientReturningLogWatch(LogWatch delegate, KubernetesClient client, KubernetesApplicationManagerContext context) {
+            this.delegate = delegate;
+            this.client = client;
+            this.context = context;
+        }
+        @Override
+        public void close() {
+            if (!closed) {
+                try {
+                    delegate.close();
+                } finally {
+                    KubernetesClientPool.returnClient(context, client);
+                    closed = true;
+                }
+            }
+        }
+        @Override
+        public java.io.InputStream getOutput() {
+            return delegate.getOutput();
+        }
+        @Override
+        public java.io.InputStream getError() {
+            return delegate.getError();
+        }
+    }

		KubernetesClient client1 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);
		KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);

	KubernetesClient client2 = (KubernetesClient) borrowObjectMethod.invoke(clusterClientPool);
	borrowObjectMethod.invoke(clusterClientPool);

[DSIP-17122][SERVER] Kubernetes Connection Pool Implementation #17510

[DSIP-17122][SERVER] Kubernetes Connection Pool Implementation #17510

Uh oh!

Conversation

Cat-Drink commented Sep 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Purpose of the pull request

Brief change log

Verify this pull request

Automated Testing

Manual Verification Steps

Pull Request Notice

Uh oh!

boring-cyborg bot commented Sep 15, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

SbloodyS left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Cat-Drink commented Oct 26, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

SbloodyS commented Oct 28, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Key Changes

Reviewed Changes

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 30, 2025

Cat-Drink commented Sep 15, 2025 •

edited

Loading

SbloodyS left a comment •

edited

Loading