[GEODE-8527] A distributed message can continue to wait for a member for which it failed to send the message

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.14.0
• Component/s: regions
• Assignee: Unassigned
• Reporter: Anilkumar Gingade
• Created: 23/Sep/20 21:08
• Updated: 23/Sep/20 21:33

Description

While trying to send/replicate a message (cache operation) by calling DistibutedCacheOperationMessage._distribute(); if there is any exception in sending the message to one of the recipient, the message processor created to wait for the replies can end up waiting for a reply from failed member.

This is observed while doing a code walk through.

The _distribute() method does keep track of nodes for which it fails to send the message; but is not using it to update the reply process created.

Probable solution:
1. Update the reply processor to remove the failed member from waiting member list
2. Handle the cache operation to address any data replication issue because of this.