Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1117948

Summary:	Members can miss the rebalance cancellation on coordinator change
Product:	[JBoss] JBoss Data Grid 6	Reporter:	Dan Berindei <dberinde>
Component:	Infinispan	Assignee:	Tristan Tarrant <ttarrant>
Status:	CLOSED CURRENTRELEASE	QA Contact:	Martin Gencur <mgencur>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	6.3.0	CC:	afield, jdg-bugs
Target Milestone:	CR3
Target Release:	6.3.0
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2015-01-26 14:06:03 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	1104045

Description Dan Berindei 2014-07-09 16:55:44 UTC

The new coordinator sends first a CH_UPDATE command to cancel the existing rebalance, and then a REBALANCE_START command to start a new rebalance. But the CH_UPDATE command is sent asynchronously, so it's possible for some members to receive it after the REBALANCE_START command.

If that happens, that node will assume that it will receive the segments it requested for the previous rebalance. But with the bug 1116969/ISPN-4484 fix, the provider node cancels the outbound transfer tasks when receiving a CH_UPDATE without a pendingCH, so the state requestor will never receive its segments.

Even without the bug 1116969/ISPN-4484 fix this is a problem, although less obvious. Between the provider node receiving the CH_UPDATE and the REBALANCE_START commands, it won't have the requestor in its write CH, so the requestor can miss transactions.

Comment 2 Alan Field 2014-07-15 11:37:32 UTC

Executed the elasticity test in Hyperion 3 times without a failure, and the resilience test 5 times without a failure with JDG 6.3.0 CR4. VERIFIED