Switching alias causing the downtime and requests are failing for fraction of second

[chandan_mahajan]

NEST/Elasticsearch.Net version : 6.3.*

Elasticsearch version :6.3

Description of the problem including expected versus actual behavior :
Changing alias from old to new index causing very little downtime probably < 1 sec.
We are repointing alias to the new index with documents loaded in it. While performing this action we keep on simulating users using jMeter. During the process of alias switching, we see a downtime of < 1 sec and request fails to get documents.

Steps to reproduce :

1. Perform alias switching. Keep on running jmeter to simulate load of 10 concurrent users.
2. Delete the old index after alias points to new index.

Pointing alias to new index :
await _clientNest.AliasAsync(ba => ba .Remove(remove => remove.Alias(aliasName).Index("*")).Add(add => add.Alias(aliasName).Index("test_new")));

We are deleting index after alias repointing succeeds.

We also tried
await _clientNest.AliasAsync(ba => ba .Remove(remove => remove.Alias(aliasName).Index("test_old")).Add(add => add.Alias(aliasName).Index("test_new"))); but still fail to get documents for fraction of second.