[QPID-4348] HA Use independent sequence numbers for identifying messages.

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.18
• Component/s: C++ Clustering
• Assignee: Alan Conway
• Reporter: Alan Conway
• Created: 27/Sep/12 16:31
• Updated: 06/Sep/16 16:23
• Resolved: 17/Jun/13 14:21

Description

Currently HA code uses queue sequence numbers to identify messasges.
This assumes that message sequence is identical on primary and backup.

Implementing new features (for example transactions) requires that we tolerate
ordering differences between primary and backups.

We introduces a new, queue-scoped HA sequence number managed by the HA plugin.
This is used to identify messages independently of queue sequence numbers.

Needs careful implementation to avoid bad performance effects.

• Avoid adding headers, use the same mechansim as we have now for sending HA seq. nos.
• Use an efficient index struct (e.g. dequeue) to map HA seq. to queue seq.

The benefits:

• required to support TX
• required for re-use of stored messages for catchup
• more robust - does not require primary and backups to have identical message ordering on their queues.