Hi @laks,
Per Transactions: Production Considerations (Sharded Clusters) this is an expected error:
On a sharded cluster, transactions that span multiple shards will error and abort if any involved shard contains an arbiter.
Arbiters do not write data, so they can cause significant operational challenges for use cases that rely on majority acknowledgement of writes. If your transaction use case requires updating documents across multiple shards, you will have to replace any arbiters with secondaries.
I strongly recommend avoiding arbiters in production as they introduce significant operational challenges. For more background, see Replica set with 3 DB Nodes and 1 Arbiter - #8 by Stennie_X.
Regards,
Stennie