[FLINK-39984][runtime][webUI] Dispatch thread dump to ioExecutor and support SAFE/FULL modes

docs/layouts/shortcodes/generated/cluster_configuration.html

@@ -62,6 +62,12 @@
             <td>Duration</td>
             <td>The shutdown timeout for cluster services like executors.</td>
         </tr>
+        <tr>
+            <td><h5>cluster.thread-dump.default-mode</h5></td>
+            <td style="word-wrap: break-word;">"FULL"</td>
+            <td>String</td>
+            <td>Default granularity of the JobManager/TaskManager thread-dump REST endpoint when no explicit <code class="highlighter-rouge">mode</code> query parameter is supplied. Allowed values (case-insensitive): <code class="highlighter-rouge">SAFE</code> (stack traces only, negligible JVM pause) and <code class="highlighter-rouge">FULL</code> (adds locked monitors and j.u.c. synchronizers, equivalent to jstack -l; pauses the JVM in a safepoint for a duration that scales with heap size and thread count -- seconds on large TaskManagers). The default is <code class="highlighter-rouge">FULL</code> to preserve historical behavior; operators of large clusters are strongly encouraged to switch to <code class="highlighter-rouge">SAFE</code> to avoid heartbeat timeouts caused by long safepoint pauses.</td>
+        </tr>
         <tr>
             <td><h5>cluster.thread-dump.stacktrace-max-depth</h5></td>
             <td style="word-wrap: break-word;">50</td>

docs/layouts/shortcodes/generated/expert_cluster_section.html

@@ -26,6 +26,12 @@
             <td>Boolean</td>
             <td>Whether processes should halt on fatal errors instead of performing a graceful shutdown. In some environments (e.g. Java 8 with the G1 garbage collector), a regular graceful shutdown can lead to a JVM deadlock. See <a href="https://issues.apache.org/jira/browse/FLINK-16510">FLINK-16510</a> for details.</td>
         </tr>
+        <tr>
+            <td><h5>cluster.thread-dump.default-mode</h5></td>
+            <td style="word-wrap: break-word;">"FULL"</td>
+            <td>String</td>
+            <td>Default granularity of the JobManager/TaskManager thread-dump REST endpoint when no explicit <code class="highlighter-rouge">mode</code> query parameter is supplied. Allowed values (case-insensitive): <code class="highlighter-rouge">SAFE</code> (stack traces only, negligible JVM pause) and <code class="highlighter-rouge">FULL</code> (adds locked monitors and j.u.c. synchronizers, equivalent to jstack -l; pauses the JVM in a safepoint for a duration that scales with heap size and thread count -- seconds on large TaskManagers). The default is <code class="highlighter-rouge">FULL</code> to preserve historical behavior; operators of large clusters are strongly encouraged to switch to <code class="highlighter-rouge">SAFE</code> to avoid heartbeat timeouts caused by long safepoint pauses.</td>
+        </tr>
         <tr>
             <td><h5>cluster.thread-dump.stacktrace-max-depth</h5></td>
             <td style="word-wrap: break-word;">50</td>

docs/layouts/shortcodes/generated/rest_v1_dispatcher.html

@@ -1383,6 +1383,16 @@
     <tr>
       <td colspan="2">Returns the thread dump of the JobManager.</td>
     </tr>
+    <tr>
+      <td colspan="2">Query parameters</td>
+    </tr>
+    <tr>
+      <td colspan="2">
+        <ul>
+<li><code>mode</code> (optional): Controls how much lock information is collected. Supported values: [SAFE, FULL]. When omitted, cluster.thread-dump.default-mode is used.</li>
+        </ul>
+      </td>
+    </tr>
     <tr>
       <td colspan="2">
         <label>
@@ -7121,6 +7131,16 @@
         </ul>
       </td>
     </tr>
+    <tr>
+      <td colspan="2">Query parameters</td>
+    </tr>
+    <tr>
+      <td colspan="2">
+        <ul>
+<li><code>mode</code> (optional): Controls how much lock information is collected. Supported values: [SAFE, FULL]. When omitted, cluster.thread-dump.default-mode is used.</li>
+        </ul>
+      </td>
+    </tr>
     <tr>
       <td colspan="2">
         <label>

docs/static/generated/rest_v1_dispatcher.yml

@@ -457,6 +457,15 @@ paths:
     get:
       description: Returns the thread dump of the JobManager.
       operationId: getJobManagerThreadDump
+      parameters:
+      - name: mode
+        in: query
+        description: "Controls how much lock information is collected. Supported values:\
+          \ [SAFE, FULL]. When omitted, cluster.thread-dump.default-mode is used."
+        required: false
+        style: form
+        schema:
+          $ref: "#/components/schemas/ThreadDumpMode"
       responses:
         "200":
           description: The request was successful.
@@ -1863,6 +1872,14 @@ paths:
         required: true
         schema:
           $ref: "#/components/schemas/ResourceID"
+      - name: mode
+        in: query
+        description: "Controls how much lock information is collected. Supported values:\
+          \ [SAFE, FULL]. When omitted, cluster.thread-dump.default-mode is used."
+        required: false
+        style: form
+        schema:
+          $ref: "#/components/schemas/ThreadDumpMode"
       responses:
         "200":
           description: The request was successful.
@@ -3911,6 +3928,11 @@ components:
           type: array
           items:
             $ref: "#/components/schemas/ThreadInfo"
+    ThreadDumpMode:
+      type: string
+      enum:
+      - SAFE
+      - FULL
     ThreadInfo:
       type: object
       properties:

flink-core/src/main/java/org/apache/flink/configuration/ClusterOptions.java

@@ -140,6 +140,31 @@ public class ClusterOptions {
                     .withDescription(
                             "The maximum stacktrace depth of TaskManager and JobManager's thread dump web-frontend displayed.");
 
+    @Documentation.Section(Documentation.Sections.EXPERT_CLUSTER)
+    public static final ConfigOption<String> THREAD_DUMP_DEFAULT_MODE =
+            key("cluster.thread-dump.default-mode")
+                    .stringType()
+                    .defaultValue("FULL")
+                    .withDescription(
+                            Description.builder()
+                                    .text(
+                                            "Default granularity of the JobManager/TaskManager thread-dump REST endpoint "
+                                                    + "when no explicit %s query parameter is supplied. ",
+                                            code("mode"))
+                                    .text(
+                                            "Allowed values (case-insensitive): %s (stack traces only, negligible JVM pause) "
+                                                    + "and %s (adds locked monitors and j.u.c. synchronizers, equivalent "
+                                                    + "to jstack -l; pauses the JVM in a safepoint for a duration that "
+                                                    + "scales with heap size and thread count -- seconds on large "
+                                                    + "TaskManagers). ",
+                                            code("SAFE"), code("FULL"))
+                                    .text(
+                                            "The default is %s to preserve historical behavior; operators of large "
+                                                    + "clusters are strongly encouraged to switch to %s to avoid "
+                                                    + "heartbeat timeouts caused by long safepoint pauses.",
+                                            code("FULL"), code("SAFE"))
+                                    .build());
+
     @Documentation.Section(Documentation.Sections.EXPERT_CLUSTER)
     public static final ConfigOption<UncaughtExceptionHandleMode> UNCAUGHT_EXCEPTION_HANDLING =
             ConfigOptions.key("cluster.uncaught-exception-handling")

flink-runtime-web/src/test/resources/rest_api_v1.snapshot

@@ -1085,7 +1085,10 @@
       "pathParameters" : [ ]
     },
     "query-parameters" : {
-      "queryParameters" : [ ]
+      "queryParameters" : [ {
+        "key" : "mode",
+        "mandatory" : false
+      } ]
     },
     "request" : {
       "type" : "object",
@@ -5429,7 +5432,10 @@
       } ]
     },
     "query-parameters" : {
-      "queryParameters" : [ ]
+      "queryParameters" : [ {
+        "key" : "mode",
+        "mandatory" : false
+      } ]
     },
     "request" : {
       "type" : "object",

flink-runtime-web/web-dashboard/src/app/pages/job-manager/thread-dump/job-manager-thread-dump.component.html

@@ -22,8 +22,34 @@
   [ngModel]="dump"
   [nzEditorOption]="editorOptions"
 ></nz-code-editor>
-<flink-addon-compact
-  [downloadHref]="downloadUrl"
-  [downloadName]="downloadName"
-  (reload)="reload()"
-></flink-addon-compact>
+<div class="thread-dump-toolbar">
+  <nz-space-compact class="thread-dump-mode-group">
+    <a
+      nz-button
+      nzSize="small"
+      [nzType]="mode === 'safe' ? 'primary' : 'default'"
+      nz-tooltip
+      nzTooltipTitle="Stack traces only, no lock info (jstack without -l). Negligible JVM pause."
+      (click)="selectMode('safe')"
+    >
+      Safe
+    </a>
+    <a
+      nz-button
+      nzSize="small"
+      [nzType]="mode === 'full' ? 'primary' : 'default'"
+      [nzDanger]="true"
+      nz-tooltip
+      nzTooltipTitle="Adds locked monitors and j.u.c. synchronizers (jstack -l). Pauses the JVM in a safepoint for a duration that scales with heap size and thread count — seconds on large JMs."
+      (click)="selectMode('full')"
+    >
+      Full
+      <i nz-icon nzType="exclamation-circle" nzTheme="fill" class="thread-dump-mode-warn"></i>
+    </a>
+  </nz-space-compact>
+  <flink-addon-compact
+    [downloadHref]="downloadUrl"
+    [downloadName]="downloadName"
+    (reload)="reload()"
+  ></flink-addon-compact>
+</div>

flink-runtime-web/web-dashboard/src/app/pages/job-manager/thread-dump/job-manager-thread-dump.component.less

@@ -28,3 +28,25 @@
     left: 0;
   }
 }
+
+.thread-dump-toolbar {
+  position: absolute;
+  top: 8px;
+  right: 32px;
+  z-index: 1;
+  display: flex;
+  gap: 8px;
+  align-items: center;
+
+  flink-addon-compact {
+    position: static;
+    top: auto;
+    right: auto;
+  }
+}
+
+.thread-dump-mode-group {
+  .thread-dump-mode-warn {
+    margin-left: 4px;
+  }
+}

flink-runtime-web/web-dashboard/src/app/pages/job-manager/thread-dump/job-manager-thread-dump.component.ts

@@ -29,21 +29,39 @@ import {
   JobManagerModuleConfig
 } from '@flink-runtime-web/pages/job-manager/job-manager.config';
 import { ConfigService, JobManagerService } from '@flink-runtime-web/services';
+import { NzButtonModule } from 'ng-zorro-antd/button';
 import { NzCodeEditorModule, EditorOptions } from 'ng-zorro-antd/code-editor';
+import { NzIconModule } from 'ng-zorro-antd/icon';
+import { NzSpaceModule } from 'ng-zorro-antd/space';
+import { NzTooltipModule } from 'ng-zorro-antd/tooltip';
+
+/** See ThreadDumpMode in task-manager-thread-dump.component.ts for semantics. */
+type ThreadDumpMode = 'safe' | 'full' | undefined;
 
 @Component({
   selector: 'flink-job-manager-thread-dump',
   templateUrl: './job-manager-thread-dump.component.html',
   styleUrls: ['./job-manager-thread-dump.component.less'],
   changeDetection: ChangeDetectionStrategy.OnPush,
-  imports: [NzCodeEditorModule, AutoResizeDirective, FormsModule, AddonCompactComponent]
+  imports: [
+    NzCodeEditorModule,
+    AutoResizeDirective,
+    FormsModule,
+    AddonCompactComponent,
+    NzButtonModule,
+    NzIconModule,
+    NzSpaceModule,
+    NzTooltipModule
+  ]
 })
 export class JobManagerThreadDumpComponent implements OnInit, OnDestroy {
   public readonly downloadName = `jobmanager_thread_dump`;
   public downloadUrl: string;
   public editorOptions: EditorOptions;
   public dump = '';
   public loading = true;
+  /** See TaskManagerThreadDumpComponent#mode. */
+  public mode: ThreadDumpMode = undefined;
 
   private readonly destroy$ = new Subject<void>();
 
@@ -66,11 +84,24 @@ export class JobManagerThreadDumpComponent implements OnInit, OnDestroy {
     this.destroy$.complete();
   }
 
+  /**
+   * Switch dump mode. Does NOT auto-reload; user must press the refresh button.
+   * Download link is updated immediately so a download matches the user's selection.
+   */
+  public selectMode(mode: 'safe' | 'full'): void {
+    if (this.mode === mode) {
+      return;
+    }
+    this.mode = mode;
+    this.updateDownloadUrl();
+    this.cdr.markForCheck();
+  }
+
   public reload(): void {
     this.loading = true;
     this.cdr.markForCheck();
     this.jobManagerService
-      .loadThreadDump()
+      .loadThreadDump(this.mode)
       .pipe(
         catchError(() => of('')),
         takeUntil(this.destroy$)
@@ -81,4 +112,9 @@ export class JobManagerThreadDumpComponent implements OnInit, OnDestroy {
         this.cdr.markForCheck();
       });
   }
+
+  private updateDownloadUrl(): void {
+    const base = `${this.configService.BASE_URL}/jobmanager/thread-dump`;
+    this.downloadUrl = this.mode ? `${base}?mode=${this.mode}` : base;
+  }
 }

flink-runtime-web/web-dashboard/src/app/pages/task-manager/thread-dump/task-manager-thread-dump.component.html

@@ -23,8 +23,34 @@
   [nzEditorOption]="editorOptions"
   (nzEditorInitialized)="nzEditorInitialized($event)"
 ></nz-code-editor>
-<flink-addon-compact
-  [downloadHref]="downloadUrl"
-  [downloadName]="downloadName"
-  (reload)="reload()"
-></flink-addon-compact>
+<div class="thread-dump-toolbar">
+  <nz-space-compact class="thread-dump-mode-group">
+    <a
+      nz-button
+      nzSize="small"
+      [nzType]="mode === 'safe' ? 'primary' : 'default'"
+      nz-tooltip
+      nzTooltipTitle="Stack traces only, no lock info (jstack without -l). Negligible JVM pause."
+      (click)="selectMode('safe')"
+    >
+      Safe
+    </a>
+    <a
+      nz-button
+      nzSize="small"
+      [nzType]="mode === 'full' ? 'primary' : 'default'"
+      [nzDanger]="true"
+      nz-tooltip
+      nzTooltipTitle="Adds locked monitors and j.u.c. synchronizers (jstack -l). Pauses the JVM in a safepoint for a duration that scales with heap size and thread count — seconds on large TMs."
+      (click)="selectMode('full')"
+    >
+      Full
+      <i nz-icon nzType="exclamation-circle" nzTheme="fill" class="thread-dump-mode-warn"></i>
+    </a>
+  </nz-space-compact>
+  <flink-addon-compact
+    [downloadHref]="downloadUrl"
+    [downloadName]="downloadName"
+    (reload)="reload()"
+  ></flink-addon-compact>
+</div>

flink-runtime-web/web-dashboard/src/app/pages/task-manager/thread-dump/task-manager-thread-dump.component.less

@@ -28,3 +28,28 @@
     left: 0;
   }
 }
+
+.thread-dump-toolbar {
+  position: absolute;
+  top: 8px;
+  right: 32px;
+  z-index: 1;
+  display: flex;
+  gap: 8px;
+  align-items: center;
+
+  // Inner <flink-addon-compact> is also position:absolute by default; neutralize
+  // it so it stacks naturally to the right of the mode selector inside this flex
+  // container.
+  flink-addon-compact {
+    position: static;
+    top: auto;
+    right: auto;
+  }
+}
+
+.thread-dump-mode-group {
+  .thread-dump-mode-warn {
+    margin-left: 4px;
+  }
+}