Daemon child: failed to write token to parent in-socket: {}

Error message

Daemon child: failed to write token to parent in-socket: {}

What it means

Final handshake step: the child writes the authentication token to the parent's in-socket (its write channel). write_all failing means that socket broke before or during the write: the parent closed its read end (gave up waiting, crashed) yielding EPIPE, or the loopback connection was reset mid-write.

Source

Thrown at rust_port/crates/flow_daemon/src/daemon.rs:566

        panic!(
            "Daemon child: failed to write token to parent out-socket: {}",
            e
        )
    });
    let mut child_out_sock = TcpStream::connect(parent_in_addr).unwrap_or_else(|e| {
        panic!(
            "Daemon child: failed to connect to parent in-socket (child write end): {}",
            e
        )
    });
    child_out_sock.set_nodelay(true).unwrap_or_else(|e| {
        panic!(
            "Daemon child: failed to set TCP_NODELAY on parent in-socket: {}",
            e
        )
    });
    child_out_sock.write_all(&token).unwrap_or_else(|e| {
        panic!(
            "Daemon child: failed to write token to parent in-socket: {}",
            e
        )
    });

    exec(&entry_name, child_in_sock, child_out_sock, param_bytes);
}

pub fn close<I, O>(h: &mut Handle<I, O>) -> std::io::Result<()> {
    h.channels.0.stream.shutdown(std::net::Shutdown::Read)?;
    h.channels.1.stream.shutdown(std::net::Shutdown::Write)?;
    Ok(())
}

pub fn close_noerr<I, O>(h: &mut Handle<I, O>) {
    if let Err(e) = h.channels.0.stream.shutdown(std::net::Shutdown::Read) {
        tracing::debug!(target: "flow_daemon", "close_noerr (in): {}", e);
    }

View on GitHub (pinned to f88ac94bcf)

Solutions

Retry the command; one-off handshake races typically clear
Address load/memory pressure so the parent survives the full handshake
flow stop, then retry to start from clean daemon state
Inspect parent logs for early exit reasons if it repeats

Defensive patterns

Strategy: retry

Prevention

Retry the command; broken handshake pipes are usually one-off races
Address memory/load pressure so the parent survives the full two-socket handshake
Run flow stop before retrying to start from clean daemon state

When it happens

Trigger: Parent times out or exits after accepting but before reading the second token write; connection reset by the OS; process tree torn down by a supervisor during the handshake.

Common situations: Heavily loaded or throttled machines stretching the two-write handshake past parent timeouts; OOM killing the parent mid-handshake; cleanup scripts killing daemon trees during spawn.

Related errors

AI-assisted analysis of facebook/flow@f88ac94bcf (2026-08-20). Data as JSON: /api/errors/3f7cc8b9cc633c66. Report an issue: GitHub.