17f13e0460
Implements plan §2 unplanned node failure RF-restore flow: - When a node recovers after failure, schedule background replication - For each shard the recovered node should own, find healthy source replica - Create migration job to copy data from surviving replica to recovered node - Dual-write starts immediately so writes go to both source and recovered node Key changes: - Enhanced `on_node_recovered` to trigger RF-restore migrations - Added `compute_shard_sources_for_rf_restore` to find healthy intra-group sources - Reuses existing migration infrastructure for consistency with node addition Cross-group fallback was already implemented in scatter.rs for RF=1 groups. Closes: miroir-mkk.5 |
||
|---|---|---|
| .. | ||
| miroir-core | ||
| miroir-ctl | ||
| miroir-proxy | ||