Tuning bulk rejections and GC in my Elasticsearch cluster

I have a 6 Node 6 shards and 2 replica elastic cluster. When have heavy indexing happening and we see the requests are getting rejected. We tried increasing the bulk queue size from 250k to 1 million , refresh interval: 30 and flush threshold 2gb. The result of that is we started seeing the indexing failures. I need someone elastic expert to help me out .