Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[HUDI-5686] Fixes data loss due to rollbacks #7828

Closed
wants to merge 5 commits into from
Closed

[HUDI-5686] Fixes data loss due to rollbacks #7828

Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
49ab579
Merge pull request #1 from apache/master
pushpavanthar Feb 2, 2023
f4cebdf
avoid creating new instants for rollbacks
Feb 2, 2023
5cb99f4
Merge branch 'master' of github.com:apache/hudi into HUDI-5686
Feb 3, 2023
f84a6df
Merge branch 'master' of github.com:apache/hudi into HUDI-5686
Feb 4, 2023
2ec476d
perform rollback before initiating commit instant
Feb 4, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ...hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieTableServiceClient.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -734,7 +734,7 @@ protected List<String> getInstantsToRollback(HoodieTableMetaClient metaClient, H
@Deprecated
public boolean rollback(final String commitInstantTime, Option<HoodiePendingRollbackInfo> pendingRollbackInfo, boolean skipLocking) throws HoodieRollbackException {
LOG.info("Begin rollback of instant " + commitInstantTime);
final String rollbackInstantTime = pendingRollbackInfo.map(entry -> entry.getRollbackInstant().getTimestamp()).orElse(HoodieActiveTimeline.createNewInstantTime());
final String rollbackInstantTime = pendingRollbackInfo.map(entry -> entry.getRollbackInstant().getTimestamp()).orElse(commitInstantTime);
final Timer.Context timerContext = this.metrics.getRollbackCtx();
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm confused, why rollback the instant itself under the current transaction.

Copy link
Author

@pushpavanthar pushpavanthar Feb 3, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi @danny0405, i just realised that this way of picking timestamp from instant itself will make features like rollbackToInstant fail. This approach fails to place rollback instant in chronological order in timeline.
I'll update the PR with below approach,
use approach similar to below method, which actually takes care of rollbacks first before creating instant for the commit.
org.apache.hudi.client.BaseHoodieWriteClient#startCommit(java.lang.String, org.apache.hudi.common.table.HoodieTableMetaClient)

try {
HoodieTable table = createTable(config, hadoopConf);
Expand Down