[opt](scheduler) Improve Graceful Shutdown Behavior for BE and FE, and Optimize Query Retry During BE Shutdown #56601
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
morningman
requested review from
dataroaring,
gavinchou and
w41ter
as code owners
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
run buildall |
TPC-DS: Total hot run time: 190772 ms |
ClickBench: Total hot run time: 31.27 s |
Contributor
FE UT Coverage ReportIncrement line coverage |
Contributor
FE Regression Coverage ReportIncrement line coverage |
morningman
force-pushed
the
be_shutdown_notice
branch
from
34da79c to
3279622
Compare
Contributor
Author
|
run buildall |
TPC-DS: Total hot run time: 189841 ms |
ClickBench: Total hot run time: 30.44 s |
Contributor
FE Regression Coverage ReportIncrement line coverage |
morningman
force-pushed
the
be_shutdown_notice
branch
2 times, most recently
from
b9eff70 to
f6b418d
Compare
Contributor
Author
|
run buildall |
morningman
force-pushed
the
be_shutdown_notice
branch
from
f6b418d to
1acd70c
Compare
There was a problem hiding this comment.
Pull Request Overview
This pull request implements graceful shutdown mechanisms for both Frontend (FE) and Backend (BE) servers, and refactors error handling for cloud retry logic. The key changes include adding server readiness checks, graceful shutdown hooks, and consolidating retry error handling into a centralized method.
- Adds graceful shutdown for FE and BE servers with wait logic for running queries/tasks
- Implements server readiness health checks that return appropriate status during startup/shutdown
- Refactors cloud retry error checking into a centralized
SystemInfoService.needRetryWithReplan()method - Moves backend shutdown state from
AtomicBooleanto serializedBackendStatus.isShutdownfield
Reviewed Changes
Copilot reviewed 20 out of 20 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| SystemInfoService.java | Adds centralized retry error detection method and constant set |
| Backend.java | Refactors setters to return boolean indicating state change; moves isShutDown to BackendStatus |
| StmtExecutor.java | Updates retry logic to use centralized error detection method |
| OlapInsertExecutor.java | Updates retry logic to use centralized error detection method |
| AbstractInsertExecutor.java | Updates retry logic to use centralized error detection method |
| StreamLoadHandler.java | Changes backend filter from isAlive to isLoadAvailable |
| MTMVTask.java | Updates retry logic to use centralized error detection method |
| RestApiStatusCode.java | Adds SERVICE_UNAVAILABLE status code |
| HealthAction.java | Adds server readiness check to health endpoint |
| ResponseEntityBuilder.java | Adds serviceUnavailable response builder method |
| HttpServer.java | Implements graceful shutdown and disables automatic shutdown hook |
| FeConstants.java | Removes unused constants including CLOUD_RETRY_E230 |
| CloudReplica.java | Updates backend filtering to use isQueryAvailable |
| DorisFE.java | Implements graceful shutdown logic and server readiness tracking |
| doris_main.cpp | Sets server readiness flags during startup/shutdown |
| exec_env.h | Adds k_is_server_ready global flag |
| exec_env.cpp | Adds post-delay sleep after graceful shutdown |
| health_action.cpp | Updates health check to return status based on server readiness |
| config.h | Adds grace_shutdown_post_delay_seconds configuration |
| config.cpp | Defines default value for grace_shutdown_post_delay_seconds |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| this.backendStatus.lastStreamLoadTime = lastStreamLoadTime; | ||
| } | ||
|
|
||
| // ATTN: This method only the value of "isQueryDisabled", |
There was a problem hiding this comment.
Incomplete sentence in comment. Should be 'This method only returns the value of "isQueryDisabled"'.
Suggested change
| // ATTN: This method only the value of "isQueryDisabled", | |
| // ATTN: This method only returns the value of "isQueryDisabled", |
| return false; | ||
| } | ||
|
|
||
| // ATTN: This method only the value of "isLoadDisabled", |
There was a problem hiding this comment.
Incomplete sentence in comment. Should be 'This method only returns the value of "isLoadDisabled"'.
Suggested change
| // ATTN: This method only the value of "isLoadDisabled", | |
| // ATTN: This method only returns the value of "isLoadDisabled", |
| @SerializedName("isActive") | ||
| public volatile boolean isActive = true; | ||
| @SerializedName("isShutdown") | ||
| public volatile boolean isShutdown = true; |
There was a problem hiding this comment.
The default value of isShutdown should be false, not true. A newly created backend should not be marked as shutdown by default.
Suggested change
| public volatile boolean isShutdown = true; | |
| public volatile boolean isShutdown = false; |
| boolean isNeedRetry = false; | ||
| if (Config.isCloudMode()) { | ||
| // cloud mode retry | ||
| // cloud mode retryF |
There was a problem hiding this comment.
Typo in comment: 'retryF' should be 'retry'.
Suggested change
| // cloud mode retryF | |
| // cloud mode retry |
| while (true) { | ||
| Thread.sleep(2000); | ||
| } | ||
| LOG.info("Doris FE main loop exited, shutting down gracefully..."); |
There was a problem hiding this comment.
This log statement is unreachable because the while loop on line 264 is infinite. The code will never reach line 267.
| LOG.info("Doris FE main loop exited, shutting down gracefully..."); | ||
| } catch (Throwable e) { | ||
| // Some exception may thrown before LOG is inited. | ||
| // Some exception may throw before LOG is inited. |
There was a problem hiding this comment.
Grammatically incorrect: 'may throw' should be 'may be thrown'.
Suggested change
| // Some exception may throw before LOG is inited. | |
| // Some exception may be thrown before LOG is inited. |
morningman
changed the title
Contributor
Author
|
run buildall |
yiguolei
previously approved these changes
Nov 5, 2025
Contributor
|
PR approved by at least one committer and no changes requested. |
github-actions
bot
added
approved
morningman
force-pushed
the
be_shutdown_notice
branch
from
5a5ae36 to
33e99f5
Compare
Contributor
Author
|
run buildall |
gavinchou
approved these changes
Nov 7, 2025
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
TPC-DS: Total hot run time: 189127 ms |
ClickBench: Total hot run time: 27.62 s |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
FE Regression Coverage ReportIncrement line coverage |
CalvinKirs
approved these changes
Nov 7, 2025
morningman
added a commit
to morningman/doris
that referenced
this pull request
Nov 7, 2025
…d Optimize Query Retry During BE Shutdown (apache#56601) Related PR: apache#23865 This PR includes the following main changes: 1. New BE Parameter: `grace_shutdown_post_delay_seconds` When using the BE graceful stop feature, after the main process waits for all currently running tasks to complete, it will continue to wait for an additional period to ensure that queries still running on other nodes have also finished. Since a BE node cannot detect the execution status of tasks on other BE nodes, this threshold may need to be increased to allow a longer waiting time. 2. Enhanced BE `api/health` Endpoint * When the BE has not yet fully started or is in the process of shutting down, the endpoint will return: * Message: `"Server is not available"` * HTTP Code: `200` * Under normal circumstances: * Message: `"OK"` * HTTP Code: `200` When using `stop_fe.sh --grace`, the FE will wait for currently running queries to finish before exiting. Note, Currently, only query tasks are waited for; import and other types of tasks are not yet included. In cloud mode, when encountering the error `"No backend available as scan node"`, the FE will now internally retry the query to reassign it to other available BE nodes.
morningman
added a commit
that referenced
this pull request
Nov 15, 2025
wyxxxcat
pushed a commit
to wyxxxcat/doris
that referenced
this pull request
Nov 18, 2025
…d Optimize Query Retry During BE Shutdown (apache#56601) ### What problem does this PR solve? Related PR: apache#23865 This PR includes the following main changes: #### BE Graceful Shutdown Improvements 1. New BE Parameter: `grace_shutdown_post_delay_seconds` When using the BE graceful stop feature, after the main process waits for all currently running tasks to complete, it will continue to wait for an additional period to ensure that queries still running on other nodes have also finished. Since a BE node cannot detect the execution status of tasks on other BE nodes, this threshold may need to be increased to allow a longer waiting time. 2. Enhanced BE `api/health` Endpoint * When the BE has not yet fully started or is in the process of shutting down, the endpoint will return: * Message: `"Server is not available"` * HTTP Code: `200` * Under normal circumstances: * Message: `"OK"` * HTTP Code: `200` #### Added FE Graceful Shutdown Support When using `stop_fe.sh --grace`, the FE will wait for currently running queries to finish before exiting. Note, Currently, only query tasks are waited for; import and other types of tasks are not yet included. #### Query Retry Optimization During BE Shutdown In cloud mode, when encountering the error `"No backend available as scan node"`, the FE will now internally retry the query to reassign it to other available BE nodes.
morningman
added a commit
to morningman/doris
that referenced
this pull request
Nov 29, 2025
…d Optimize Query Retry During BE Shutdown (apache#56601) ### What problem does this PR solve? Related PR: apache#23865 This PR includes the following main changes: #### BE Graceful Shutdown Improvements 1. New BE Parameter: `grace_shutdown_post_delay_seconds` When using the BE graceful stop feature, after the main process waits for all currently running tasks to complete, it will continue to wait for an additional period to ensure that queries still running on other nodes have also finished. Since a BE node cannot detect the execution status of tasks on other BE nodes, this threshold may need to be increased to allow a longer waiting time. 2. Enhanced BE `api/health` Endpoint * When the BE has not yet fully started or is in the process of shutting down, the endpoint will return: * Message: `"Server is not available"` * HTTP Code: `200` * Under normal circumstances: * Message: `"OK"` * HTTP Code: `200` #### Added FE Graceful Shutdown Support When using `stop_fe.sh --grace`, the FE will wait for currently running queries to finish before exiting. Note, Currently, only query tasks are waited for; import and other types of tasks are not yet included. #### Query Retry Optimization During BE Shutdown In cloud mode, when encountering the error `"No backend available as scan node"`, the FE will now internally retry the query to reassign it to other available BE nodes.
morningman
added a commit
to morningman/doris
that referenced
this pull request
Nov 29, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
9 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What problem does this PR solve?
Related PR: #23865
This PR includes the following main changes:
BE Graceful Shutdown Improvements
New BE Parameter:
grace_shutdown_post_delay_secondsWhen using the BE graceful stop feature, after the main process waits for all currently running tasks to complete, it will continue to wait for an additional period to ensure that queries still running on other nodes have also finished.
Since a BE node cannot detect the execution status of tasks on other BE nodes, this threshold may need to be increased to allow a longer waiting time.
Enhanced BE
api/healthEndpointWhen the BE has not yet fully started or is in the process of shutting down, the endpoint will return:
"Server is not available"200Under normal circumstances:
"OK"200Added FE Graceful Shutdown Support
When using
stop_fe.sh --grace, the FE will wait for currently running queries to finish before exiting.Note, Currently, only query tasks are waited for; import and other types of tasks are not yet included.
Query Retry Optimization During BE Shutdown
In cloud mode, when encountering the error
"No backend available as scan node",the FE will now internally retry the query to reassign it to other available BE nodes.
Release note
None
Check List (For Author)
Test
select sleep(30)sh bin/stop_fe.sh --graceselectfinishselect sleep(30)is executed, FE will also wait this query finishBehavior changed:
Does this need documentation?
Check List (For Reviewer who merge this PR)