We have previously added protection to reduce QoS impact upon a node rejoining (#96521), after being down for a short period of time and requiring to be caught up on any missed messages.
This protection was known to be insufficient to prevent all QoS impact but was sufficient to reduce the impact.
RACv2 operating in apply_to_all provides the prevention, but isn't planned to be enabled by default until v25.1 (currently in process of cutting 24.3).
This issue is to enhance the existing protection to provide further reduction in QoS impact.
We have previously added protection to reduce QoS impact upon a node rejoining (#96521), after being down for a short period of time and requiring to be caught up on any missed messages.
This protection was known to be insufficient to prevent all QoS impact but was sufficient to reduce the impact.
RACv2 operating in
apply_to_all
provides the prevention, but isn't planned to be enabled by default until v25.1 (currently in process of cutting 24.3).This issue is to enhance the existing protection to provide further reduction in QoS impact.
See investigation in #132615.
Jira issue: CRDB-43281