Denis Chudov created IGNITE-27610:
-------------------------------------
Summary: Trigger transaction recovery proactively when coordinator
leaves a cluster
Key: IGNITE-27610
URL: https://issues.apache.org/jira/browse/IGNITE-27610
Project: Ignite
Issue Type: Bug
Reporter: Denis Chudov
As for now, transaction recovery is triggered lazily during write intent
resolution, when write intent of another transaction is detected.
This may cause lots of hanging tx cleanup futures when nodes leave and enter
the cluster multiple times under the load, leading to heap occupation and gc
pauses.
*Definition of done*
Transaction recovery is triggered when transaction's coordinator leaves
cluster. This should be background process with proper throttling similar to
transaction cleanup after node restart.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
