[Opt](scanner-scheduler) Opt scanner scheduler starvation issue. #40641
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
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
kaka11chen
force-pushed
the
opt_scanner_schedule_starvation_1
branch
from
86c08f6 to
ca2e7e3
Compare
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
kaka11chen
force-pushed
the
opt_scanner_schedule_starvation_1
branch
from
ca2e7e3 to
af7b099
Compare
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 42132 ms |
TPC-DS: Total hot run time: 194081 ms |
|
TeamCity be ut coverage result: |
ClickBench: Total hot run time: 32.69 s |
yiguolei
reviewed
Sep 23, 2024
kaka11chen
force-pushed
the
opt_scanner_schedule_starvation_1
branch
3 times, most recently
from
a1b1e78 to
2087b56
Compare
Contributor
Author
|
run buildall |
Contributor
|
PR approved by anyone and no changes requested. |
kaka11chen
force-pushed
the
opt_scanner_schedule_starvation_1
branch
from
2087b56 to
725afd0
Compare
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 43020 ms |
TPC-DS: Total hot run time: 195427 ms |
ClickBench: Total hot run time: 33.62 s |
|
TeamCity be ut coverage result: |
morningman
approved these changes
Sep 26, 2024
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
kaka11chen
added a commit
to kaka11chen/doris
that referenced
this pull request
Sep 30, 2024
…che#40641) When a scanner scheduler is stuck in executing a scan task, other scan tasks will starve and have no chance to execute, which will affect other queries. Currently, the scan task hopes to scan as much data as possible to reduce the overhead of scheduling switching. Currently, it hopes to obtain up to 10MB of data in `doris_scanner_row_bytes`. However, if a query scans a table with many rows of data, but the filtering rate is very high, the filter will eventually filter out a lot of data and will never get 10MB of data. It will keep getting and executing expression filtering, which will cause other scan tasks to starve. The current solution is to check `max_run_time_ms` by `MonotonicStopWatch`. After executing for a maximum of 1s, it will yield self's task for other tasks. When the scan task executes some time-consuming tasks, it needs to slice to do it.
kaka11chen
added a commit
to kaka11chen/doris
that referenced
this pull request
Sep 30, 2024
…che#40641) When a scanner scheduler is stuck in executing a scan task, other scan tasks will starve and have no chance to execute, which will affect other queries. Currently, the scan task hopes to scan as much data as possible to reduce the overhead of scheduling switching. Currently, it hopes to obtain up to 10MB of data in `doris_scanner_row_bytes`. However, if a query scans a table with many rows of data, but the filtering rate is very high, the filter will eventually filter out a lot of data and will never get 10MB of data. It will keep getting and executing expression filtering, which will cause other scan tasks to starve. The current solution is to check `max_run_time_ms` by `MonotonicStopWatch`. After executing for a maximum of 1s, it will yield self's task for other tasks. When the scan task executes some time-consuming tasks, it needs to slice to do it.
morningman
pushed a commit
that referenced
this pull request
Sep 30, 2024
) ## Proposed changes Backport #40641
morningman
pushed a commit
that referenced
this pull request
Sep 30, 2024
) ## Proposed changes Backport #40641
cjj2010
pushed a commit
to cjj2010/doris
that referenced
this pull request
Oct 12, 2024
…che#40641) ## Proposed changes ### Issue When a scanner scheduler is stuck in executing a scan task, other scan tasks will starve and have no chance to execute, which will affect other queries. Currently, the scan task hopes to scan as much data as possible to reduce the overhead of scheduling switching. Currently, it hopes to obtain up to 10MB of data in `doris_scanner_row_bytes`. However, if a query scans a table with many rows of data, but the filtering rate is very high, the filter will eventually filter out a lot of data and will never get 10MB of data. It will keep getting and executing expression filtering, which will cause other scan tasks to starve. ### Solution The current solution is to check `max_run_time_ms` by `MonotonicStopWatch`. After executing for a maximum of 1s, it will yield self's task for other tasks. When the scan task executes some time-consuming tasks, it needs to slice to do it.
16 tasks
morningman
pushed a commit
that referenced
this pull request
Dec 30, 2024
…ate materialization case of parquet reader (#46121) ### What problem does this PR solve? Related PR: #40641 Problem Summary: [Fix](parquet-reader) Fixed the issue of excessive scanning data in late materialization case of parquet reader introduced by #40641 in scenarios with particularly high filtering rates.
github-actions bot
pushed a commit
that referenced
this pull request
Dec 30, 2024
…ate materialization case of parquet reader (#46121) ### What problem does this PR solve? Related PR: #40641 Problem Summary: [Fix](parquet-reader) Fixed the issue of excessive scanning data in late materialization case of parquet reader introduced by #40641 in scenarios with particularly high filtering rates.
github-actions bot
pushed a commit
that referenced
this pull request
Dec 30, 2024
…ate materialization case of parquet reader (#46121) ### What problem does this PR solve? Related PR: #40641 Problem Summary: [Fix](parquet-reader) Fixed the issue of excessive scanning data in late materialization case of parquet reader introduced by #40641 in scenarios with particularly high filtering rates.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
Proposed changes
Issue
When a scanner scheduler is stuck in executing a scan task, other scan tasks will starve and have no chance to execute, which will affect other queries. Currently, the scan task hopes to scan as much data as possible to reduce the overhead of scheduling switching. Currently, it hopes to obtain up to 10MB of data in
doris_scanner_row_bytes. However, if a query scans a table with many rows of data, but the filtering rate is very high, the filter will eventually filter out a lot of data and will never get 10MB of data. It will keep getting and executing expression filtering, which will cause other scan tasks to starve.Solution
The current solution is to check
max_run_time_msbyMonotonicStopWatch. After executing for a maximum of 1s, it will yield self's task for other tasks. When the scan task executes some time-consuming tasks, it needs to slice to do it.