How to Use Point in Time for Pagination in Elasticsearch

Use point_in_time (PIT) in Elasticsearch to create a consistent snapshot of your data for pagination. Open a PIT with POST /_search/point_in_time, then use the returned pit_id in your search queries with search_after to paginate reliably without missing or duplicating results.

Syntax

The basic syntax involves three parts:

• Open a PIT: Request a point in time snapshot to get a pit_id.
• Search with PIT: Use the pit_id in your search query along with search_after for pagination.
• Close the PIT: Optionally close the PIT to free resources.

This keeps your pagination consistent even if data changes during navigation.

json

POST /_search/point_in_time
{
  "index": "your-index-name",
  "keep_alive": "1m"
}

POST /your-index-name/_search
{
  "size": 10,
  "pit": {
    "id": "<pit_id_from_previous_response>",
    "keep_alive": "1m"
  },
  "search_after": ["<sort_value_from_last_hit>"],
  "sort": [
    {"timestamp": "asc"},
    {"_shard_doc": "asc"}
  ]
}

DELETE /_search/point_in_time
{
  "id": "<pit_id>"
}

Example

This example shows how to open a PIT, perform a paginated search using search_after, and then close the PIT.

json

POST /_search/point_in_time
{
  "index": "products",
  "keep_alive": "2m"
}

# Response:
# {
#   "pit_id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA..."
# }

POST /products/_search
{
  "size": 5,
  "pit": {
    "id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA...",
    "keep_alive": "2m"
  },
  "sort": [
    {"price": "asc"},
    {"_shard_doc": "asc"}
  ]
}

# Assume last hit sort values: [19.99, 12345]

POST /products/_search
{
  "size": 5,
  "pit": {
    "id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA...",
    "keep_alive": "2m"
  },
  "search_after": [19.99, 12345],
  "sort": [
    {"price": "asc"},
    {"_shard_doc": "asc"}
  ]
}

DELETE /_search/point_in_time
{
  "id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA..."
}

Output

{
  "pit_id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA..."
}

{
  "hits": {
    "hits": [
      {"_id": "1", "sort": [9.99, 10001]},
      {"_id": "2", "sort": [14.99, 10002]},
      {"_id": "3", "sort": [19.99, 12345]},
      {"_id": "4", "sort": [24.99, 10004]},
      {"_id": "5", "sort": [29.99, 10005]}
    ]
  }
}

Common Pitfalls

• Not using search_after with PIT: This causes repeated results or missing pages.
• Not sorting on unique fields: Always include a tiebreaker like _shard_doc to ensure stable sorting.
• Forgetting to keep PIT alive: PIT expires after keep_alive time; extend it if pagination takes longer.
• Not closing PIT: While optional, closing PIT frees cluster resources.

json

POST /products/_search
{
  "size": 5,
  "pit": {
    "id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA...",
    "keep_alive": "1m"
  },
  "sort": [
    {"price": "asc"}
  ]
  # Missing _shard_doc as tiebreaker
}

# Right way:
POST /products/_search
{
  "size": 5,
  "pit": {
    "id": "DXF1ZXJ5QW5kRmV0Y2gBAAAAAAA...",
    "keep_alive": "1m"
  },
  "sort": [
    {"price": "asc"},
    {"_shard_doc": "asc"}
  ]
}

Quick Reference

Point in Time (PIT) Pagination Cheat Sheet:

Step	Action	Notes
1	Open PIT	Use POST /_search/point_in_time with index and keep_alive
2	Search with PIT	Include pit_id and search_after with stable sort
3	Paginate	Use last hit's sort values in search_after
4	Close PIT	Optional but recommended to free resources

Step	Action	Notes
1	Open PIT	Use POST /_search/point_in_time with index and keep_alive
2	Search with PIT	Include pit_id and search_after with stable sort
3	Paginate	Use last hit's sort values in search_after
4	Close PIT	Optional but recommended to free resources

Key Takeaways

Use point in time (PIT) to get a consistent snapshot for pagination in Elasticsearch.

Always combine PIT with search_after and a stable sort including a tiebreaker like _shard_doc.

Set and extend keep_alive to keep the PIT valid during pagination.

Close the PIT when done to free cluster resources.

Avoid using from/size for deep pagination; PIT with search_after is more efficient.