HBase service check failed consistently because the HBase Master died as it was trying to write the WAL and could not find enough good datanodes. The service check for HBase was run before the service check for HDFS. There may be a situation in which we're not giving enough time for all the DN's to come up before the HBase service check is being performed, and if it would make more sense to service check HDFS before we service check HBase due to dependencies.
With that said it feels like the RU service check order should be:
- Everything else