risingwavelabs/risingwave · error · StreamExecutorError

RpcError

Error message

RpcError

What it means

This is the RpcError variant of StreamExecutorError, which transparently wraps an RpcError from risingwave_rpc_client. Stream executors throw it when a gRPC call to another service (meta, compute, or another worker) fails. Since the variant uses #[error(transparent)] and #[from], the underlying RPC error's Display and source are passed through unchanged.

Solutions

1. Check connectivity between the compute node and the meta service (address, port, DNS).
2. Inspect the wrapped RpcError source (via its `source()`) for the concrete gRPC status (Unavailable, DeadlineExceeded, etc.).
3. Retry the operation if the failure was transient; the stream framework typically restarts the actor.
4. Verify service versions are compatible (proto mismatch can cause RPC failures).

Example fix

// before
let resp = rpc_client.do_something().await?; // RpcError propagates as executor error
// after
let resp = rpc_client.do_something().await.inspect_err(|e| {
    tracing::warn!(error = ?e, "rpc failed, will be retried by actor restart");
})?;

// before issuing RPC: verify endpoint reachable
async fn ensure_rpc_ready(addr: &str) -> anyhow::Result<()> {
    tokio::net::TcpStream::connect(addr).await
        .map_err(|e| anyhow::anyhow!("rpc endpoint {addr} unreachable: {e}"))?;
    Ok(())
}

fn as_rpc_error(e: &StreamExecutorError) -> Option<&risingwave_rpc_client::error::RpcError> {
    use risingwave_rpc_client::error::RpcError;
    e.0.downcast_ref::<RpcError>()
}

match result {
    Err(e) if matches!(e.variant_name(), "RpcError") => {
        // inspect inner RpcError, apply backoff and retry
    }
    other => other.map(|_| ())?,
}

When it happens

Trigger: Any executor path that performs an RPC via risingwave_rpc_client and uses `?` on a Result whose error is RpcError, e.g. remote sink/table lookups or meta node communication during streaming. The `#[from]` impl converts the RpcError automatically.

Common situations: Meta node down or unreachable; network partitions between compute nodes; RPC deadlines exceeded under load; service restarting during deployment.

Related errors

• RpcError
• bigquery insert error: end of resp stream
• Cannot open client to compute node
• end of response stream
• ExchangeChannelClosed

AI-assisted analysis of risingwavelabs/risingwave@6469eb736d (2026-09-11). Data as JSON: /api/errors/60eb3a9fecdba1a3. Report an issue: GitHub.

Appendix: source

Thrown at src/stream/src/executor/error.rs:78

    ),

    // TODO: remove this after state table is fully used
    #[error("Serialize/deserialize error: {0}")]
    SerdeError(
        #[source]
        #[backtrace]
        BoxedError,
    ),

    #[error("Sink error: sink_id={1}, error: {0}")]
    SinkError(
        #[source]
        #[backtrace]
        SinkError,
        SinkId,
    ),

    #[error(transparent)]
    RpcError(
        #[from]
        #[backtrace]
        RpcError,
    ),

    #[error("Channel closed: {0}")]
    ChannelClosed(String),

    #[error(transparent)]
    ExchangeChannelClosed(
        #[from]
        #[backtrace]
        ExchangeChannelClosed,
    ),

    #[error("Failed to align barrier: expected `{0:?}` but got `{1:?}`")]
    AlignBarrier(Box<Barrier>, Box<Barrier>),

View on GitHub (pinned to 6469eb736d)