[IGNITE-10771] Print troubleshooting hint when exchange latch got stucked

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.5
• Fix Version/s: 2.8
• Component/s: cache
• Labels: usability
• Assignee: Pavel Kovalenko
• Reporter: Pavel Kovalenko
• Created: 20/Dec/18 19:39
• Updated: 15/Oct/19 15:02
• Resolved: 15/Oct/19 15:02

Description

Sometimes users face with a problem when exchange latch can't be completed:

2018-12-12 07:07:57:563 [exchange-worker-#42] WARN 
o.a.i.i.p.c.d.d.p.GridDhtPartitionsExchangeFuture:488 - Unable to await 
partitions release latch within timeout: ClientLatch 
[coordinator=ZookeeperClusterNode [id=6b9fc6e4-5b6a-4a98-be4d-6bc1aa5c014c, 
addrs=[172.17.0.1, 10.0.230.117, 0:0:0:0:0:0:0:1%lo, 127.0.0.1], order=3, 
loc=false, client=false], ackSent=true, super=CompletableLatch [id=exchange, 
topVer=AffinityTopologyVersion [topVer=45, minorTopVer=1]]] 

It may indicate that some node in a cluster can' t finish partitions release (finish all ongoing operations at the previous topology version) or it can be silent network problem.
We should print to log a hint how to troubleshoot it to reduce the number of questions about such problem.

Issue Links

• links to: GitHub Pull Request #6968