apache/cassandra · error · InvalidRequestException

Unknown node

Error message

Unknown node 

What it means

INSERT/UPDATE (apply) into the virtual table uses the partition key as a numeric NodeId. ClusterMetadata.directory has no endpoint registered for that id, so the write cannot be routed and InvalidRequestException is thrown.

Solutions

1. Look up valid node ids (e.g. from the corresponding system virtual table or nodetool status) and use one of those as the partition key.
2. If targeting a restarted cluster, re-fetch node ids — ids are not stable across topology changes.
3. Verify the target node is still a member: ClusterMetadata.current().directory should contain it.
4. Correct the integer value; node ids are internal, not user-chosen.

Example fix

// before
INSERT INTO vt (node_id, ...) VALUES (42, ...);   // 42 not in cluster
// after
INSERT INTO vt (node_id, ...) VALUES (1, ...);    // id fetched from directory/status

int nodeId = ...;
InetAddressAndPort ep = ClusterMetadata.current().directory.endpoint(new NodeId(nodeId));
if (ep == null) throw new IllegalArgumentException("Unknown node " + nodeId + "; fetch valid ids from the directory");

try { session.execute(write); }
catch (InvalidRequestException e) {
    if (e.getMessage().startsWith("Unknown node"))
        // refresh node id list and retry with a valid id
}

When it happens

Trigger: Writing to the virtual table with a partition key that is not a currently registered node id (typo, stale id of a removed node, wrong integer).

Common situations: Scripts caching node ids across cluster restarts; writes targeting a node after it was decommissioned/removenode'd; manual experimentation with arbitrary ids.

Understand the failure class

Background: 'Could not be found', 'does not exist', 'not found in database': the resource-not-found family when an ID, slug, key, or URI lookup comes back empty — this error's family across 20 libraries.

Related errors

• Can only unset '" + name + "'
• Cannot filter this table by partial partition key
• Cannot update '" + name + "'
• metadata + " currently only supports querying single…
• Modification is not supported by table " + metadata

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/73a5128d6fddc165. Report an issue: GitHub.

Appendix: source

Thrown at src/java/org/apache/cassandra/db/virtual/RemoteToLocalVirtualTable.java:501

    private static void addExpressions(RowFilter rowFilter, List<ColumnMetadata> cms, int start, ByteBuffer[] values, Operator op, Operator lastOp)
    {
        for (int i = start ; i < values.length ; ++i)
            rowFilter.add(cms.get(i), i + 1 == values.length ? lastOp : op, values[i]);
    }

    private static ClusteringIndexSliceFilter filter(TableMetadata metadata, ClusteringBound<?> start, ClusteringBound<?> end, boolean reversed)
    {
        return new ClusteringIndexSliceFilter(Slices.with(metadata.comparator, Slice.make(start, end)), reversed);
    }

    @Override
    public void apply(PartitionUpdate update)
    {
        int nodeId = Int32Type.instance.compose(update.partitionKey().getKey());
        InetAddressAndPort endpoint = ClusterMetadata.current().directory.endpoint(new NodeId(nodeId));
        if (endpoint == null)
            throw new InvalidRequestException("Unknown node " + nodeId);

        DeletionInfo deletionInfo = update.deletionInfo();
        if (!deletionInfo.getPartitionDeletion().isLive())
        {
            truncate(endpoint).syncThrowUncheckedOnInterrupt();
            return;
        }

        int pkCount = local.partitionKeyColumns().size();
        ByteBuffer[] pkBuffer, ckBuffer;
        {
            int ckCount = local.clusteringColumns().size();
            pkBuffer = pkCount == 1 ? null : new ByteBuffer[pkCount];
            ckBuffer = new ByteBuffer[ckCount];
        }

        PartitionUpdate.Builder builder = null;
        ArrayDeque<Promise<Void>> results = new ArrayDeque<>();

View on GitHub (pinned to 88fd0f6a0e)