GC warnings and Time out exceptions on bulk update

It looks like you're simply pushing data quicker than what your current cluster can handle. Have you tried reducing the batch size? I don't think 100k batches gives better performance than, say, 10k.