I run multiple instances of ES on the same system to obtain greater memory utilization (ES 1.7.5). Each instance has a different name (think host1-1, host1-2). Somehow, an additional instance of one of the nodes got started. The side effect of this is that ES decided to move data to that additional instance. When the cluster was restarted, it went red because there were only two instances. I've been able to manually recover either by manually moving the indices from the new node directory or by re-starting an extra third instance and relocating away from it.
Are there any practices I can use to prevent this from happening in the future?