Today the DryRun is opportunistic and assumes the latest activity stats come from all nodes at the same time. The point-in-time cluster view needs to be more consistent.
- Per-node freshness vs stale-node policy
- gap detection
And somehow it needs to deal with changing shape of the whole PostgreSQL cluster deployment.
Today the DryRun is opportunistic and assumes the latest activity stats come from all nodes at the same time. The point-in-time cluster view needs to be more consistent.
And somehow it needs to deal with changing shape of the whole PostgreSQL cluster deployment.