Partition Management

Partitions allow you to organize data within a collection for better performance and management. This guide covers all partition operations.

Creating Partitions

Create a new partition:

await client.createPartition({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
});

Checking Partition Existence

Check if a partition exists:

const exists = await client.hasPartition({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
});

console.log('Partition exists:', exists.value);

Listing Partitions

List all partitions in a collection:

const result = await client.listPartitions({
  collection_name: 'my_collection',
});

console.log('Partitions:', result.partition_names);
// Output: ['_default', 'partition_1', 'partition_2']

Partition Statistics

Get statistics for a partition:

const stats = await client.getPartitionStatistics({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
});

console.log('Row count:', stats.row_count);
console.log('Statistics:', stats.stats);

Partition Loading

Load Partitions

Load partitions into memory:

// Asynchronous load
await client.loadPartitions({
  collection_name: 'my_collection',
  partition_names: ['partition_1', 'partition_2'],
});

// Synchronous load (waits until loaded)
await client.loadPartitionsSync({
  collection_name: 'my_collection',
  partition_names: ['partition_1', 'partition_2'],
});

Release Partitions

Release partitions from memory:

await client.releasePartitions({
  collection_name: 'my_collection',
  partition_names: ['partition_1', 'partition_2'],
});

Dropping Partitions

Drop a partition (this permanently deletes all data in the partition):

await client.dropPartition({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
});

Working with Partitions

Inserting Data into Partitions

Insert data into a specific partition:

await client.insert({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
  data: [
    {
      vector: [/* ... */],
      text: 'data for partition 1',
    },
  ],
});

Querying Partitions

Query data from specific partitions:

const results = await client.query({
  collection_name: 'my_collection',
  expr: 'id > 100',
  partition_names: ['partition_1'],
  output_fields: ['vector', 'text'],
});

Searching Partitions

Search within specific partitions:

const results = await client.search({
  collection_name: 'my_collection',
  data: [/* vector */],
  partition_names: ['partition_1', 'partition_2'],
  limit: 10,
});

Default Partition

Every collection has a default partition named _default. Data inserted without specifying a partition goes to the default partition:

// Data goes to _default partition
await client.insert({
  collection_name: 'my_collection',
  data: [/* ... */],
});

// Explicitly specify partition
await client.insert({
  collection_name: 'my_collection',
  partition_name: 'partition_1',
  data: [/* ... */],
});

Partition Best Practices

1. Use partitions for data organization: Organize data by time, category, or region
2. Load only needed partitions: Load only the partitions you need to query
3. Partition naming: Use descriptive names that reflect the partition’s purpose
4. Default partition: Use the default partition for general data
5. Partition size: Keep partitions at reasonable sizes for optimal performance

Next Steps

• Learn about Collection Management
• Explore Data Operations
• Check out Best Practices