[ https://issues.apache.org/jira/browse/IGNITE-9147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16568171#comment-16568171 ]
Alexei Scherbakov commented on IGNITE-9147: ------------------------------------------- Root cause is race between forced tx rollback and tx map phase. > When server node left cluster on high load, cluster take hang on > PartitionalExchange > ------------------------------------------------------------------------------------ > > Key: IGNITE-9147 > URL: https://issues.apache.org/jira/browse/IGNITE-9147 > Project: Ignite > Issue Type: Bug > Components: general > Affects Versions: 2.5 > Reporter: ARomantsov > Assignee: Alexei Scherbakov > Priority: Critical > Fix For: 2.7 > > > I ran a simple test > 1) Start 15 servers node > 2) Start client with long transaction > 3) Additional start 5 client with loading in many caches (near 2 thousand) > 4) Stop 1 server node, wait 1 minute and start it back > Cluster freenze on more than hour, then license end -- This message was sent by Atlassian JIRA (v7.6.3#76005)