Spark Streaming: How to handle failure in Spark when connecting to multiple kafka cluster via Union of Dstream?

Question

I have a requirement where I have to read from multiple Kafka clusters (more than 20 clusters) via spark streaming.

I am able to read all of them by basically creating kafka Direct Stream for all the kafka cluster and performing union on streams.

JavaDStream kafkaStream = null;
for (String kafkaPod : nextgenKafkaPods) {
        JavaInputDStream kStream = null;
        try {
            kStream = createKafkaStream(appContext.getJavaStreamingContext(), kafkaParamsList.get(kafkaPod));
            if (kafkaStream == null) {
                kafkaStream = kStream;
            } else {
                kafkaStream = kafkaStream.union(kStream);
            }
        } catch (Exception e) {
            LOGGER.error("Unable to create dStream for pod:" + kafkaPod, e);
        }

    }

It works fine until/unless one of the stream gets some failure (error can come in between or at start only like ssl handshake error). In such scenario, even if only 1 or few kafka cluster are having issue still whole spark job goes down. How to prevent failure in such scenarios?