[feature](cloud-compaction) Support shadow tablet to do cumulative compaction during schema change in cloud mode #39558
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. |
Collaborator
Author
|
run buildall |
…oud mode (apache#37293) In cloud mode, when do schema change, shadow tablet encounters -235 because it cant do cumulative compaction in the case of a large number of loads. And it will prevents the user from continuing to loads. Implementation details: 1. When start schema change, record the end convert rowset version `alter_version` into SchemaChangeJob. 2. For origin tablet, only can do base compaction in [0, `alter_version`] and do cumulative compaction in (`alter_version`, N]. can not do compaction across `alter_verison` such as compaction [a, `alter_version` + n]. 3. For shadow tablet, cannot do base compaction and and do cumulative compaction in (`alter_version`, N]. 4. When the schema change failed because FE or BE coredump, it will retry. When retry the schema change, it will get the `alter_version` from meta_serive, and continue to do it. 5. When finish the schema change job or cancel it, we need to clear the schema change job. Before this pr, it will cover by next schema change.
ff30024 to
a909702
Compare
Collaborator
Author
|
run buildall |
TPC-H: Total hot run time: 38124 ms |
TPC-DS: Total hot run time: 196619 ms |
ClickBench: Total hot run time: 31.64 s |
Collaborator
Author
|
run cloud_p0 |
Collaborator
Author
|
run cloud_p0 |
1 similar comment
Collaborator
Author
|
run cloud_p0 |
Hastyshell
commented
Aug 20, 2024
Comment on lines
155
to
156
| << compaction.input_versions(0) | ||
| << " input_version_end=" << compaction.input_versions(1) |
Collaborator
Author
There was a problem hiding this comment.
Here is supicious
Collaborator
Author
|
run buildall |
Collaborator
Author
|
run buildall |
TPC-H: Total hot run time: 39740 ms |
TPC-H: Total hot run time: 38069 ms |
65244ae to
3ea46ed
Compare
Collaborator
Author
|
run buildall |
TPC-H: Total hot run time: 38659 ms |
TPC-DS: Total hot run time: 191408 ms |
ClickBench: Total hot run time: 30.59 s |
Collaborator
Author
|
run cloud_p0 |
gavinchou
previously approved these changes
Aug 22, 2024
Collaborator
Author
|
run external |
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
Collaborator
Author
|
run buildall |
| } | ||
|
|
||
| Status CloudSchemaChangeJob::_convert_historical_rowsets(const SchemaChangeParams& sc_params) { | ||
| Status CloudSchemaChangeJob::_convert_historical_rowsets(const SchemaChangeParams& sc_params, |
Contributor
There was a problem hiding this comment.
warning: function '_convert_historical_rowsets' exceeds recommended size/complexity thresholds [readability-function-size]
Status CloudSchemaChangeJob::_convert_historical_rowsets(const SchemaChangeParams& sc_params,
^Additional context
be/src/cloud/cloud_schema_change_job.cpp:222: 177 lines including whitespace and comments (threshold 80)
Status CloudSchemaChangeJob::_convert_historical_rowsets(const SchemaChangeParams& sc_params,
^
gavinchou
approved these changes
Aug 23, 2024
Contributor
|
PR approved by at least one committer and no changes requested. |
TPC-H: Total hot run time: 38256 ms |
TPC-DS: Total hot run time: 186022 ms |
ClickBench: Total hot run time: 30.55 s |
luwei16
approved these changes
Aug 24, 2024
gavinchou
pushed a commit
that referenced
this pull request
Sep 11, 2024
…mpaction during schema change in cloud mode (#39558) In cloud mode, when do schema change, shadow tablet encounters -235 because it cant do cumulative compaction in the case of a large number of loads. And it will prevents the user from continuing to loads. Implementation details: 1. When start schema change, record the end convert rowset version `alter_version` into SchemaChangeJob. 2. For origin tablet, only can do base compaction in [0, `alter_version`] and do cumulative compaction in (`alter_version`, N]. can not do compaction across `alter_verison` such as compaction [a, `alter_version` + n]. 3. For shadow tablet, cannot do base compaction and and do cumulative compaction in (`alter_version`, N]. 4. When the schema change failed because FE or BE coredump, it will retry. When retry the schema change, it will get the `alter_version` from meta_serive, and continue to do it. 5. When finish the schema change job or cancel it, we need to clear the schema change job. Before this pr, it will cover by next schema change. co-author(main author): @Lchangliang original PR: #37293 --------- Co-authored-by: Lightman <31928846+Lchangliang@users.noreply.github.com>
Closed
16 tasks
dataroaring
pushed a commit
that referenced
this pull request
Mar 13, 2025
…n=true` (#48399) ### What problem does this PR solve? #39558 add a config to support shadow tablet to do cumulative compaction during schema change in cloud mode to avoid -235 error on new tablet in the case of a large number of loads. However, this introduces correctness problem on merge-on-write table because some rowsets' delete bitmaps are wrong if there are cumu compactions on new tablet when SC calculate delete bitmaps for incremental rowsets after converting historical data. This PR introduce a new type of compaction `STOP_TOKEN` to fail all existing compaction jobs and disallow doing any compaction on tablet and change the SC process as following: 1. converting historical data 2. register stop token on new tablet to fail all existing compaction jobs and disallow doing any compaction on new tablet 3. calculate delete bitmap for incremental rowsets without lock 4. calculate delete bitmap for incremental rowsets with lock 5. commit SC job and remove stop token ---- ref: #29386
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Mar 13, 2025
…n=true` (apache#48399) apache#39558 add a config to support shadow tablet to do cumulative compaction during schema change in cloud mode to avoid -235 error on new tablet in the case of a large number of loads. However, this introduces correctness problem on merge-on-write table because some rowsets' delete bitmaps are wrong if there are cumu compactions on new tablet when SC calculate delete bitmaps for incremental rowsets after converting historical data. This PR introduce a new type of compaction `STOP_TOKEN` to fail all existing compaction jobs and disallow doing any compaction on tablet and change the SC process as following: 1. converting historical data 2. register stop token on new tablet to fail all existing compaction jobs and disallow doing any compaction on new tablet 3. calculate delete bitmap for incremental rowsets without lock 4. calculate delete bitmap for incremental rowsets with lock 5. commit SC job and remove stop token ---- ref: apache#29386
koarz
pushed a commit
to koarz/doris
that referenced
this pull request
Jun 4, 2025
…n=true` (apache#48399) ### What problem does this PR solve? apache#39558 add a config to support shadow tablet to do cumulative compaction during schema change in cloud mode to avoid -235 error on new tablet in the case of a large number of loads. However, this introduces correctness problem on merge-on-write table because some rowsets' delete bitmaps are wrong if there are cumu compactions on new tablet when SC calculate delete bitmaps for incremental rowsets after converting historical data. This PR introduce a new type of compaction `STOP_TOKEN` to fail all existing compaction jobs and disallow doing any compaction on tablet and change the SC process as following: 1. converting historical data 2. register stop token on new tablet to fail all existing compaction jobs and disallow doing any compaction on new tablet 3. calculate delete bitmap for incremental rowsets without lock 4. calculate delete bitmap for incremental rowsets with lock 5. commit SC job and remove stop token ---- ref: apache#29386
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 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
In cloud mode, when do schema change, shadow tablet encounters -235 because it cant do cumulative compaction in the case of a large number of loads. And it will prevents the user from continuing to loads. Implementation details:
alter_versioninto SchemaChangeJob.alter_version] and do cumulative compaction in (alter_version, N]. can not do compaction acrossalter_verisonsuch as compaction [a,alter_version+ n].alter_version, N].alter_versionfrom meta_serive, and continue to do it.co-author(main author): @Lchangliang
original PR: #37293