This is an automated email from the ASF dual-hosted git repository.
panyuepeng pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/flink.git
The following commit(s) were added to refs/heads/master by this push:
new 4684e3e541a [FLINK-38716][docs] Describe the impact of
slot.request.max-interval on balanced tasks scheduling. (#27270)
4684e3e541a is described below
commit 4684e3e541a379393051fa1504afc41d2bc2bd2a
Author: Yuepeng Pan <[email protected]>
AuthorDate: Wed Apr 8 11:37:07 2026 +0800
[FLINK-38716][docs] Describe the impact of slot.request.max-interval on
balanced tasks scheduling. (#27270)
Co-authored-by: davidradl <[email protected]>
---
.../tasks-scheduling/balanced_tasks_scheduling.md | 12 ++++++++++++
.../tasks-scheduling/balanced_tasks_scheduling.md | 16 ++++++++++++++++
2 files changed, 28 insertions(+)
diff --git
a/docs/content.zh/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
b/docs/content.zh/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
index 034a2c431a0..2ce8c1e92e8 100644
---
a/docs/content.zh/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
+++
b/docs/content.zh/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
@@ -100,6 +100,18 @@ Slot 资源准备就绪后：
- `taskmanager.load-balance.mode`: `tasks`
+<span class="label label-info">注意</span> 在故障切换场景中，
+当资源被释放并处理资源请求时，资源视图的延迟更新可能导致分配结果未达到最优平衡。
+在这种情况下，可通过适当增大
+[`slot.request.max-interval`]({{< ref "docs/deployment/config"
>}}#slot-request-max-interval) 的数值来改善此问题。
+例如，每次可尝试以 `50` 毫秒为增量进行调整。
+提高该数值将使作业调度期间的槽位请求与可用资源视图更加稳定，从而使任务能够尽可能均衡分配。
+但相应地，这会延长任务调度阶段的整体时长。
+因此，增大此配置值也会增加 [`slot.request.timeout`]({{< ref "docs/deployment/config"
>}}#slot-request-timeout) 参数超时的风险。
+值得强调的是，若在充分提高该选项数值后，仍出现非最优的任务均衡现象，
+可在 <a href="https://issues.apache.org/jira/browse/FLINK-38715";>FLINK-38715</a>
中报告此问题，
+报告中应包含调度相关配置的说明及观察到的具体现象描述。
+
<a name="more-details"></a>
## 更多详情
diff --git
a/docs/content/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
b/docs/content/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
index 7006cc56e0c..ea9ef5bac9c 100644
--- a/docs/content/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
+++ b/docs/content/docs/deployment/tasks-scheduling/balanced_tasks_scheduling.md
@@ -116,6 +116,22 @@ You can enable balanced tasks scheduling through the
following configuration ite
- `taskmanager.load-balance.mode`: `tasks`
+<span class="label label-info">Note</span> During failover scenarios,
+when resources are released and resource requests are processed,
+the delayed updates in the resource view may lead to allocation results that
are not optimally balanced.
+In such cases, you can improve the situation by appropriately increasing the
value of
+[`slot.request.max-interval`]({{< ref "docs/deployment/config"
>}}#slot-request-max-interval).
+For example, you can try increasing it by `50` milliseconds each time for
adjustment.
+Raising this value will make the slot requests and available resource view
more stable during job scheduling,
+allowing tasks to be allocated as balanced as possible.
+However, this will correspondingly extend the overall duration of the task
scheduling phase.
+Therefore, increasing this configuration value will also raise the risk of
+[`slot.request.timeout`]({{< ref "docs/deployment/config"
>}}#slot-request-timeout) parameter timeouts.
+It must be emphasized that if non-best task balancing still occurs after the
value of
+option has been sufficiently increased, you may report it in
+<a href="https://issues.apache.org/jira/browse/FLINK-38715";>FLINK-38715</a>,
+which should include descriptions of scheduling-related configurations and the
observed phenomena.
+
## More details
See the <a href="https://cwiki.apache.org/confluence/x/U56zDw";>FLIP-370</a>
for more details.
