I assume you mean shards of new indexes?
When an index is created elasticsearch makes sure that shards that were created on a newer version node will only ever be allocated on other nodes that have a version at least as up-to-date as the one where is was created. So depending on where a shard was created it might never be able to relocate anywhere else and also replicas will not be assigned on lower version nodes. Upgrading the other nodes will fix this in this case.
In addition, it looks like data2 and data3 are running low on disk space which might be another reason for the decision to allocate the shards on the new node in the beginning.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.