[IGNITE-12930] DistributedProcess fails node if unable to send single message to coordinator

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 2.9
• Assignee: Nikita Amelchev
• Reporter: Maxim Muzafarov
• Created: 21/Apr/20 22:38
• Updated: 24/Jul/20 07:07
• Resolved: 23/Jul/20 15:57

Description

The DistributedProcess fails the local node (FailureHandler CRITICAL_ERROR thrown) if unable to send a message to the coordinator (e.g. the coordinator fails right before the single message is sent).

            try {
                ctx.io().sendToGridTopic(p.crdId, GridTopic.TOPIC_DISTRIBUTED_PROCESS, singleMsg, SYSTEM_POOL);
            }
            catch (IgniteCheckedException e) {
                log.error("Unable to send message to coordinator.", e);

                ctx.failure().process(new FailureContext(CRITICAL_ERROR,
                    new Exception("Unable to send message to coordinator.", e)));
            }

Expected behaviour

If the ClusterTopologyCheckedException occurs need to wait for the NODE_LEFT event of the coordinator node and re-init the distributed process future.

Issue Links

• links to: GitHub Pull Request #7714
• links to: PR#7714