[
https://issues.apache.org/jira/browse/IGNITE-26621?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Scherbakov updated IGNITE-26621:
---------------------------------------
Description:
Currently implicit getAll/containsAll transactions from a client are send to a
coordinator node in a proxy mode.
This prevents scaling due to additional network overhead.
Instead we should:
# Keep old behavior for a single partition batch
# Send batches in direct mode on all nodes to improve parallelism
# Execute each batch as independent explicit transaction
# Wait for all batches completion and release locks
# Retry on lock conflict
was:
Currently implicit batches from a client are send to a coordinator node in a
proxy mode.
This prevents scaling due to additional network overhead.
Instead we can try:
# Keep old behavior for single partition batch
# Send batches in direct mode on all nodes to improve parallelism
# Execute each batch as independent explicit transaction
# Wait for all
> Improve implicit getAll/containsAll handling from a client
> ----------------------------------------------------------
>
> Key: IGNITE-26621
> URL: https://issues.apache.org/jira/browse/IGNITE-26621
> Project: Ignite
> Issue Type: Improvement
> Reporter: Alexey Scherbakov
> Assignee: Alexey Scherbakov
> Priority: Major
> Labels: ignite-3
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Currently implicit getAll/containsAll transactions from a client are send to
> a coordinator node in a proxy mode.
> This prevents scaling due to additional network overhead.
> Instead we should:
> # Keep old behavior for a single partition batch
> # Send batches in direct mode on all nodes to improve parallelism
> # Execute each batch as independent explicit transaction
> # Wait for all batches completion and release locks
> # Retry on lock conflict
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
