How do search_after and PIT support deep pagination?

Question

IT Interview Practice · Accepted Answer

Use `search_after` to continue after the last result of the previous page. Use a PIT, a point-in-time search view, to keep the dataset stable while moving through pages.

## Open a view and read the first page

In Kibana Dev Tools, open a PIT:

```http
POST /products/_pit?keep_alive=1m
```

Copy the returned `id` into the placeholder below. Assume `price` is a numeric field; the lifetime and page size are illustrative:

```http
POST /_search
{
  "size": 2,
  "pit": {"id":"PIT_ID_FROM_RESPONSE","keep_alive":"1m"},
  "query": {"match_all":{}},
  "sort": [{"price":"asc"},{"_shard_doc":"asc"}]
}
```

Suppose the last hit contains `"sort": [120, 17]`. The price is 120; the second value breaks ties between hits with the same price inside this PIT.

## Use that exact array for the next page

```http
POST /_search
{
  "size": 2,
  "pit": {"id":"LATEST_PIT_ID","keep_alive":"1m"},
  "query": {"match_all":{}},
  "sort": [{"price":"asc"},{"_shard_doc":"asc"}],
  "search_after": [120, 17]
}
```

Replace the placeholder with the latest returned PIT ID. Keep the query and sort unchanged, and copy all sort values from the actual last hit rather than calculating them. Close the PIT when done; an expired PIT requires a defined restart strategy because a new view may contain different data. This suits sequential navigation, not jumping directly to an arbitrary page number.

[Official documentation](https://www.elastic.co/docs/reference/elasticsearch/rest-apis/paginate-search-results).

Open a view and read the first page

Use that exact array for the next page

Related questions