[IGNITE-3418] TcpDiscoverySpi: lots of duplicated node failed messages

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.6
• Fix Version/s: 1.7
• Component/s: general
• Assignee: Unassigned
• Reporter: Semen Boikov
• Created: 05/Jul/16 04:11
• Updated: 29/Nov/18 17:07
• Resolved: 06/Jul/16 14:11

Description

Currently in TcpDiscoverySpi there is safety check called when discovery worker thread is idle: if node from 'failed list' still in ring then send 'node failed' message for such node ('ServerImpl.checkFailedNodesList'). This can lead to lots of duplicated unnecessary messages.
This check should be changed: need send event only once and only in case when node which marked node as failed left cluster.

Issue Links

• relates to: IGNITE-10198 test cases fail in TcpDiscoveryMultiThreadedTest: testCustomEventOnJoinCoordinatorStop and testClientContinuousQueryCoordinatorStop
• relates to: IGNITE-10249 TcpDiscoveryMultiThreadedTest#testCustomEventNodeRestart: Getting affinity for topology version earlier than affinity is calculated