It’s a topology change that impacts transactions rollback, not vice verse. I 
would wait while more experienced Apache committers who maintain caching 
components will chime in. They should be able to get to the bottom.

In the meanwhile please attach the logs from all the nodes (servers and 
clients, that are running on your application side). Hope you preserved them.

—
Denis

> On Sep 16, 2016, at 9:09 AM, yfernando <yohan.ferna...@tudor.com> wrote:
> 
> No, the node that failed was a server node.
> 
> About the rollback, yes indeed. A few times that grid has hung, we have seen
> a similar lock on rollback. Why would a transaction rollback impact the
> topology?
> 
> This thread dump was taken at least 10 minutes after the node died so in an
> ideal world, the grid should have recovered from the topology change which
> occurred by the node going down.
> 
> 
> 
> 
> 
> --
> View this message in context: 
> http://apache-ignite-users.70518.x6.nabble.com/One-failing-node-stalling-the-whole-cluster-tp5372p7805.html
> Sent from the Apache Ignite Users mailing list archive at Nabble.com.

Reply via email to