Cluster shard allocation filtering doesn't affect primary shards

Elastic Stack Elasticsearch
1

On ES 2.4.3, applying cluster shard filtering such as the one below only moves replica shards to other nodes. Primary ones got stuck on the node. Is this a bug or by design?

PUT /_cluster/settings
{
"transient" : {
"cluster.routing.allocation.exclude._ip" : "10.0.0.1"
}
}

2

found out it's due to rolling upgrade from 2.4.3 to 2.4.4