[IGNITE-1334] Issues with concurrent node stop and cache stop

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: ignite-1.4
• Component/s: cache
• Assignee: Unassigned
• Reporter: Semen Boikov
• Created: 31/Aug/15 09:24
• Updated: 01/Sep/15 06:00
• Resolved: 01/Sep/15 06:00

Description

There are two issue when cache is destroyed concurrently with node stop:
1. onKernalStop can be called twice for the same cache, it causes hang (for example GridCacheQueryManager.onKernalStop - second attempt to block busyLock which is already blocked will hang).
2. If some exchanges are still in progress GridDhtPartitionsExchangeFutures are completed with error, I see that at least initFut listener created in the 'onReceive(final UUID nodeId, final GridDhtPartitionsFullMessage msg)' does not check initFut result.

Sub-Tasks

• Synchronous cache operation can deadlock on async semaphore.: Open