risingwavelabs/risingwave · error · MetaError

fragment id from actor is different from fragment

Error message

fragment id {} from actor {} is different from fragment {}

What it means

compose_fragment validates that every actor passed for a fragment actually declares the same fragment_id as the fragment being composed. If any actor's `fragment_id` disagrees with the enclosing `fragment_id`, the metadata would be internally inconsistent (an actor attached to the wrong fragment), so the controller bails out before persisting the fragment.

Solutions

1. Check where the actors vector was built and ensure each actor.fragment_id is set from the same fragment being composed
2. Log the offending actor_id and fragment_id from the message to find which grouping step produced the mismatch
3. If this follows an upgrade or manual meta-store edits, verify the streaming_job/actor rows in the meta DB are consistent
4. Report a bug to RisingWave if the mismatch arises from stock plan creation, since callers are internal

Example fix

// before
let actors = build_actors(&plan); // actors may span fragments
// after
let actors: Vec<_> = all_actors.into_iter().filter(|a| a.fragment_id == fragment_id).collect();

assert!(actors.iter().all(|a| a.fragment_id == fragment_id), "actor/fragment id mismatch");

fn all_actors_match_fragment(actors: &[Actor], fragment_id: FragmentId) -> bool { actors.iter().all(|a| a.fragment_id == fragment_id) }

When it happens

Trigger: Calling CatalogController::compose_fragment (internally during table fragment creation/persisting, e.g. create table MV/materialized view flow) with an actors list where at least one Actor has fragment_id != the fragment's id — typically a bug in code that groups actors into fragments.

Common situations: Custom stream plan generation or patched actor distribution code that assigns actors to fragments incorrectly; internal state corruption between the stream graph builder and the fragment composer; merging fragments from an older cluster upgraded across versions where actor metadata format changed.

Understand the failure class

Background: "This is a bug, please report it": internal invariant violations, unreachable panics, and SNH errors explained — this error's family across 47 libraries.

Related errors

• cannot find StreamActor of actor
• expected NoShuffle but got
• fragment missing in shared actor info map
• fragment not found in shared actor info
• fragments not found

AI-assisted analysis of risingwavelabs/risingwave@6469eb736d (2026-09-11). Data as JSON: /api/errors/f13e758b8331a05c. Report an issue: GitHub.

Appendix: source

Thrown at src/meta/src/controller/fragment.rs:427

        let stream_node = stream_node.to_protobuf();
        let mut upstream_fragments = HashSet::new();
        visit_stream_node_body(&stream_node, |body| {
            if let NodeBody::Merge(m) = body {
                assert!(
                    upstream_fragments.insert(m.upstream_fragment_id),
                    "non-duplicate upstream fragment"
                );
            }
        });

        let mut pb_actors = vec![];

        let mut pb_actor_status = HashMap::new();
        let mut pb_actor_splits = HashMap::new();

        for actor in actors {
            if actor.fragment_id != fragment_id {
                bail!(
                    "fragment id {} from actor {} is different from fragment {}",
                    actor.fragment_id,
                    actor.actor_id,
                    fragment_id
                )
            }

            let ActorInfo {
                actor_id,
                fragment_id,
                worker_id,
                splits,
                vnode_bitmap,
                expr_context,
                config_override,
                ..
            } = actor;

View on GitHub (pinned to 6469eb736d)