coordinator crashes on bad logs access

[Hennzau]

Describe the bug

When I try to view logs using dora logs, the coordinator crashes if I input incorrect information (e.g., an invalid UUID or name). The same happens when there is an error, such as a wrong node ID or when multiple dataflows have the same name.

Environment (please complete the following information):

• System: macOS
• Dora version: tag v0.3.6

[Hennzau]

The same issue occurs with some other commands (dora stop [uuid] with an invalid UUID). This is due to the coordinator control event loop.

ControlRequest::Logs { uuid, name, node } => {
    let dataflow_uuid = if let Some(uuid) = uuid {
        uuid
    } else if let Some(name) = name {
        resolve_name(name, &running_dataflows, &archived_dataflows)?
    } else {
        bail!("No UUID provided")
    };

    let reply = retrieve_logs(
        &running_dataflows,
        &archived_dataflows,
        dataflow_uuid,
        node.into(),
        &mut daemon_connections,
        clock.new_timestamp(),
    )
    .await
    .map(ControlRequestReply::Logs);
    let _ = reply_sender.send(reply);
}

I don't think we should stop the coordinator if we can't resolve the dataflow UUID. I'll wrap it so it just sends an error reply instead.

Now it looks like that:

ControlRequest::Logs { uuid, name, node } => {
let dataflow_uuid = if let Some(uuid) = uuid {
    Ok(uuid)
} else if let Some(name) = name {
    resolve_name(name, &running_dataflows, &archived_dataflows)
} else {
    Err(eyre!("No uuid"))
};

match dataflow_uuid {
    Ok(uuid) => {
        let reply = retrieve_logs(
            &running_dataflows,
            &archived_dataflows,
            uuid,
            node.into(),
            &mut daemon_connections,
            clock.new_timestamp(),
        )
        .await
        .map(ControlRequestReply::Logs);
        let _ = reply_sender.send(reply);
    }
    Err(err) => {
        let _ = reply_sender.send(Err(err));
    }
}

And in case of bad logs name/uuid it looks like that:

[ERROR]
unexpected reply to daemon logs: Error("no dataflow with name `test`")

[haixuanTao]

Yes. I agree that this should not happened. And this looks good!

Could you open a PR?

Thanks in advance :)