How to handle shard failures in ElasticSearch

I'm trying to make my application that uses ElasticSearch resilient to failures. Looking at the Java client documentation on indexing responses it says I should look for shard errors in the response and handle them.

ReplicationResponse.ShardInfo shardInfo = indexResponse.getShardInfo();
if (shardInfo.getTotal() != shardInfo.getSuccessful()) {
    // TODO: Handle the situation where number of successful shards is less than total shards
}
if (shardInfo.getFailed() > 0) {
    for (ReplicationResponse.ShardInfo.Failure failure : shardInfo.getFailures()) {
        String reason = failure.reason();
        // TODO: Handle the potential failures
    }
}

How do I handle those failure scenarios? I want to make sure that my data is safely stored in ElasticSearch before continuing. If there were no successfully shards obviously I can just retry, however if there were some successful and some failed or not all shards succeeded what should I do.

I couldn't see any guidance in the documentation around this (Java client or REST API's).