Skip to content

Own-origin recovery: a failover self catch-up entry shares the recovery range's node id on the sender #1404

Open

Description

@kriszyp

Follow-up from Recover a node's own writes from its peers after it loses its log tail, and hold record locks until then (#1383).

What happens: on the sender, a failover self catch-up subscription entry (an older receiver asking for its own origin after failover) still lowers the replay start computed from the subscription list, and it names the same node, so it shares a node id in subscribedNodeIds with an own-origin recovery range from a capable receiver (replication/replicationConnection.ts, the loop that finds the earliest start). With both present, the recovery range's per-node entry and the catch-up's start can overwrite each other.

Possible fix: treat the two as mutually exclusive on a capable receiver (drop the failover entry when a recovery range is present), or key the recovery range separately.

— Claude Opus 5.5

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Fields

    Priority

    P3

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions