I configured a index lifecycle policy. And at the end of the hot phase I configured a Searchable snapshot backed on a repository hosted on our S3 bucket. I have chosen the convert to fully mounted index option.
There is no delete phase defined in index lifecycle. I saw the index rollover happening today and a new index getting created. For a brief time the older index showed twice the size on monitoring page. Then snapshot started and completed.
What was surprising is that older index got deleted from the cluster. I was not expecting that.
More info: The index lifecycle policy earlier had a delete section when the index was intially created. Then later on I modified and removed the delete section. If the older version of policy with delete section is what was baked into the initial index created then I kind of understand the behaviour.
Now I see the deleted index back with a prefix restored-
I did search for the older data to be sure that the S3 based index was searchable.