[QPID-8288] timeout on connection that is not writable does not cleanup on primary

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: qpid-cpp-1.39.0
• Fix Version/s: qpid-cpp-1.40.0
• Component/s: C++ Broker
• Assignee: Gordon Sim
• Reporter: Gordon Sim
• Created: 25/Mar/19 18:37
• Updated: 24/Apr/19 09:05
• Resolved: 25/Mar/19 21:04

Description

If an incoming socket from a backup is not writable at the point it times out, then the sessions get cleaned up but the connection observers are not notified of the closed connection (as the code used by the io layer is not yet deleted; it is only deleted when the socket is writable).

For the primary in an ha cluster this means it does not fully handle the lost connection.