[ 
https://issues.apache.org/jira/browse/IGNITE-25825?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Alexey Scherbakov updated IGNITE-25825:
---------------------------------------
    Description: 
Currently write intents are cleaned up syncronously on commit path.
This adds additional latency up to 1RTT in cross-partition scenario.
We should clean up intents and unlock locks asynchronously, as soon as commit 
record is applied:

{code:java}
return finishTransaction(enlistedPartitionGroups, txId, commit, commitTimestamp)
                .thenApply(txResult -> {
                            // Clean up and unlock asynchronously.
                            txManager.cleanup(replicationGroupId, 
enlistedPartitions, commit, commitTimestamp, txId);

                            return txResult;
                        }
                );
{code}


  was:
Currently write intents are cleaned up syncronously on commit path.
This add additional latency up to 1RTT in cross-partition scenario.
We should clean up intents and unlock locks asynchronously, as soon as commit 
record is applied:

{code:java}
return finishTransaction(enlistedPartitionGroups, txId, commit, commitTimestamp)
                .thenApply(txResult -> {
                            // Clean up and unlock asynchronously.
                            txManager.cleanup(replicationGroupId, 
enlistedPartitions, commit, commitTimestamp, txId);

                            return txResult;
                        }
                );
{code}



> Fully asynchronous write intent cleanup
> ---------------------------------------
>
>                 Key: IGNITE-25825
>                 URL: https://issues.apache.org/jira/browse/IGNITE-25825
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Vladislav Pyatkov
>            Priority: Major
>              Labels: ignite-3
>
> Currently write intents are cleaned up syncronously on commit path.
> This adds additional latency up to 1RTT in cross-partition scenario.
> We should clean up intents and unlock locks asynchronously, as soon as commit 
> record is applied:
> {code:java}
> return finishTransaction(enlistedPartitionGroups, txId, commit, 
> commitTimestamp)
>                 .thenApply(txResult -> {
>                             // Clean up and unlock asynchronously.
>                             txManager.cleanup(replicationGroupId, 
> enlistedPartitions, commit, commitTimestamp, txId);
>                             return txResult;
>                         }
>                 );
> {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to