Verify Index Integrity

ParadeDB provides amcheck-style index verification functions to detect corruption and validate the structural integrity of ParadeDB indexes. These functions are useful for:

Proactive corruption detection before issues become critical
Validating index health after hardware failures or unexpected shutdowns
Verifying backup integrity
Debugging index-related issues

Basic Verification

The pdb.verify_index function performs structural integrity checks on a ParadeDB index:

SELECT * FROM pdb.verify_index('search_idx');

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(diagnostics.verifyIndex("search_idx"));

from paradedb import paradedb_verify_index

paradedb_verify_index("search_idx")

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(engine, "search_idx")

ParadeDB.paradedb_verify_index("search_idx")

await dbContext.Database.VerifyIndex("search_idx").ToListAsync();

This returns a table with three columns:

Column	Type	Description
`check_name`	text	Name of the verification check
`passed`	boolean	Whether the check passed
`details`	text	Additional information about the check result

Example Output

               check_name               | passed |                    details
----------------------------------------+--------+-----------------------------------------------
 search_idx: schema_valid               | t      | Index schema loaded successfully
 search_idx: index_readable             | t      | Index reader opened successfully
 search_idx: checksums_valid            | t      | All segment checksums validated successfully
 search_idx: segment_metadata_valid     | t      | 3 segments validated successfully

Heap Reference Validation

To verify that all indexed entries still exist in the heap table, use the heapallindexed option:

SELECT * FROM pdb.verify_index('search_idx', heapallindexed := true);

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
  }),
);

from paradedb import paradedb_verify_index

paradedb_verify_index("search_idx", heapallindexed=True)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true
)

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions { HeapAllIndexed = true }
).ToListAsync();

This adds an additional check that validates every indexed ctid (tuple identifier) references a valid row in the table. This is particularly useful for detecting index entries that reference deleted or non-existent rows.

The heapallindexed option can be slow on large indexes as it must verify every document. Consider using sample_rate for quick spot checks on large indexes.

Sampling for Large Indexes

For large indexes, you can check a random sample of documents instead of all documents:

-- Check 10% of documents
SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    sample_rate := 0.1
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
    sampleRate: 0.1,
  }),
);

from paradedb import paradedb_verify_index

paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    sample_rate=0.1,
)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    sample_rate=0.1,
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  sample_rate: 0.1
)

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions
    {
        HeapAllIndexed = true,
        SampleRate = 0.1,
    }
).ToListAsync();

Verification Settings

Progress Reporting

For long-running verifications, enable progress reporting to see status updates:

SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    report_progress := true
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
    reportProgress: true,
  }),
);

from paradedb import paradedb_verify_index

paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    report_progress=True,
)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    report_progress=True,
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  report_progress: true
)

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions
    {
        HeapAllIndexed = true,
        ReportProgress = true,
    }
).ToListAsync();

Progress messages are emitted via Postgres’s NOTICE channel.

Verbose Mode

For detailed logging including segment-by-segment progress and resume hints, enable verbose mode:

SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    report_progress := true,
    verbose := true
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
    reportProgress: true,
    verbose: true,
  }),
);

from paradedb import paradedb_verify_index

paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    report_progress=True,
    verbose=True,
)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    report_progress=True,
    verbose=True,
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  report_progress: true,
  verbose: true
)

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions
    {
        HeapAllIndexed = true,
        ReportProgress = true,
        Verbose = true,
    }
).ToListAsync();

Stopping on First Error

To stop verification immediately when the first error is found (similar to pg_amcheck --on-error-stop):

SELECT * FROM pdb.verify_index('search_idx', on_error_stop := true);

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    onErrorStop: true,
  }),
);

from paradedb import paradedb_verify_index

paradedb_verify_index("search_idx", on_error_stop=True)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    on_error_stop=True,
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  on_error_stop: true
)

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions { OnErrorStop = true }
).ToListAsync();

Parallel Verification

A single verify_index call processes segments sequentially within one Postgres backend. For very large indexes, you can distribute verification across multiple database connections by specifying which segments each connection should check using the segment_ids parameter. This allows you to utilize multiple CPU cores by running verification in parallel processes.

Listing Segments

First, list all segments in the index:

SELECT * FROM pdb.index_segments('search_idx');

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(diagnostics.indexSegments("search_idx"));

from paradedb import paradedb_index_segments

paradedb_index_segments("search_idx")

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_index_segments(engine, "search_idx")

ParadeDB.paradedb_index_segments("search_idx")

await dbContext.Database.IndexSegments("search_idx").ToListAsync();

 partition_name | segment_idx | segment_id | num_docs | num_deleted | max_doc
----------------+-------------+------------+----------+-------------+---------
 search_idx     |           0 | b7e661af   |    10000 |           0 |   10000
 search_idx     |           1 | b4fc1b40   |    10000 |           0 |   10000
 search_idx     |           2 | 9894b412   |    10000 |           0 |   10000
 search_idx     |           3 | 4d0168d6   |     5000 |           0 |    5000

Verifying Specific Segments

Then, verify specific segments using the segment_ids parameter:

-- Worker 1: Verify even segments
SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    segment_ids := ARRAY[0, 2]
);

-- Worker 2: Verify odd segments
SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    segment_ids := ARRAY[1, 3]
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

const segments = await db.execute(diagnostics.indexSegments("search_idx"));
const evenSegments = segments
  .map((row) => row.segment_idx)
  .filter((segmentIdx) => segmentIdx % 2 === 0);
const oddSegments = segments
  .map((row) => row.segment_idx)
  .filter((segmentIdx) => segmentIdx % 2 === 1);

// Worker 1: Verify even segments
await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
    segmentIds: evenSegments,
  }),
);

// Worker 2: Verify odd segments
if (oddSegments.length) {
  await db.execute(
    diagnostics.verifyIndex("search_idx", {
      heapAllIndexed: true,
      segmentIds: oddSegments,
    }),
  );
}

from paradedb import paradedb_verify_index

# Worker 1: Verify even segments
paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    segment_ids=[0, 2],
)

# Worker 2: Verify odd segments
paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    segment_ids=[1, 3],
)

from paradedb.sqlalchemy import diagnostics

# Worker 1: Verify even segments
diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    segment_ids=[0, 2],
)

# Worker 2: Verify odd segments
diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    segment_ids=[1, 3],
)

# Worker 1: Verify even segments
ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  segment_ids: [0, 2]
)

# Worker 2: Verify odd segments
ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  segment_ids: [1, 3]
)

var segments = await dbContext.Database.IndexSegments("search_idx").ToListAsync();
var evenSegments = segments
    .Select(segment => segment.SegmentIdx)
    .Where(segmentIdx => segmentIdx % 2 == 0)
    .ToArray();
var oddSegments = segments
    .Select(segment => segment.SegmentIdx)
    .Where(segmentIdx => segmentIdx % 2 == 1)
    .ToArray();

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions
    {
        HeapAllIndexed = true,
        SegmentIds = evenSegments,
    }
).ToListAsync();

if (oddSegments.Length > 0)
{
    await dbContext.Database.VerifyIndex(
        "search_idx",
        new VerifyIndexOptions
        {
            HeapAllIndexed = true,
            SegmentIds = oddSegments,
        }
    ).ToListAsync();
}

Distributing Verification Across Workers

Distribute verification across N workers:

-- Get segments for worker 0 (of 4 workers)
SELECT array_agg(segment_idx) AS segments
FROM pdb.index_segments('search_idx')
WHERE segment_idx % 4 = 0;

-- Run verification with those segments
SELECT * FROM pdb.verify_index('search_idx',
    heapallindexed := true,
    segment_ids := (
        SELECT array_agg(segment_idx)
        FROM pdb.index_segments('search_idx')
        WHERE segment_idx % 4 = 0
    )
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

const segments = await db.execute(diagnostics.indexSegments("search_idx"));

await db.execute(
  diagnostics.verifyIndex("search_idx", {
    heapAllIndexed: true,
    segmentIds: segments
      .map((row) => row.segment_idx)
      .filter((segmentIdx) => segmentIdx % 4 === 0),
  }),
);

from paradedb import paradedb_index_segments, paradedb_verify_index

paradedb_verify_index(
    "search_idx",
    heapallindexed=True,
    segment_ids=[
        row["segment_idx"]
        for row in paradedb_index_segments("search_idx")
        if row["segment_idx"] % 4 == 0
    ],
)

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_index(
    engine,
    "search_idx",
    heapallindexed=True,
    segment_ids=[
        row["segment_idx"]
        for row in diagnostics.paradedb_index_segments(engine, "search_idx")
        if row["segment_idx"] % 4 == 0
    ],
)

ParadeDB.paradedb_verify_index(
  "search_idx",
  heapallindexed: true,
  segment_ids: ParadeDB.paradedb_index_segments("search_idx").filter_map { |row|
    row["segment_idx"] if row["segment_idx"] % 4 == 0
  })

var segments = await dbContext.Database.IndexSegments("search_idx").ToListAsync();

await dbContext.Database.VerifyIndex(
    "search_idx",
    new VerifyIndexOptions
    {
        HeapAllIndexed = true,
        SegmentIds = segments
            .Select(segment => segment.SegmentIdx)
            .Where(segmentIdx => segmentIdx % 4 == 0)
            .ToArray(),
    }
).ToListAsync();

Verifying All ParadeDB Indexes

To verify all ParadeDB indexes in the database at once:

SELECT * FROM pdb.verify_all_indexes();

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(diagnostics.verifyAllIndexes());

from paradedb import paradedb_verify_all_indexes

paradedb_verify_all_indexes()

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_verify_all_indexes(engine)

ParadeDB.paradedb_verify_all_indexes()

await dbContext.Database.VerifyAllIndexes().ToListAsync();

Filtering by Pattern

Filter indexes by schema or name pattern (using SQL LIKE syntax):

-- Verify indexes in the 'public' schema only
SELECT * FROM pdb.verify_all_indexes(schema_pattern := 'public');

-- Verify indexes matching a name pattern
SELECT * FROM pdb.verify_all_indexes(index_pattern := 'search_%');

-- Combine filters
SELECT * FROM pdb.verify_all_indexes(
    schema_pattern := 'app_%',
    index_pattern := '%_idx',
    heapallindexed := true
);

import { diagnostics } from "@paradedb/drizzle-paradedb";

// Verify indexes in the 'public' schema only
await db.execute(
  diagnostics.verifyAllIndexes({
    schemaPattern: "public",
  }),
);

// Verify indexes matching a name pattern
await db.execute(
  diagnostics.verifyAllIndexes({
    indexPattern: "search_%",
  }),
);

// Combine filters
await db.execute(
  diagnostics.verifyAllIndexes({
    schemaPattern: "public",
    indexPattern: "search_%",
    heapAllIndexed: true,
  }),
);

from paradedb import paradedb_verify_all_indexes

# Verify indexes in the 'public' schema only
paradedb_verify_all_indexes(schema_pattern="public")

# Verify indexes matching a name pattern
paradedb_verify_all_indexes(index_pattern="search_%")

# Combine filters
paradedb_verify_all_indexes(
    schema_pattern="app_%",
    index_pattern="%_idx",
    heapallindexed=True,
)

from paradedb.sqlalchemy import diagnostics

# Verify indexes in the 'public' schema only
diagnostics.paradedb_verify_all_indexes(
    engine,
    schema_pattern="public",
)

# Verify indexes matching a name pattern
diagnostics.paradedb_verify_all_indexes(
    engine,
    index_pattern="search_%",
)

# Combine filters
diagnostics.paradedb_verify_all_indexes(
    engine,
    schema_pattern="app_%",
    index_pattern="%_idx",
    heapallindexed=True,
)

# Verify indexes in the 'public' schema only
ParadeDB.paradedb_verify_all_indexes(schema_pattern: "public")

# Verify indexes matching a name pattern
ParadeDB.paradedb_verify_all_indexes(index_pattern: "search_%")

# Combine filters
ParadeDB.paradedb_verify_all_indexes(
  schema_pattern: "app_%",
  index_pattern: "%_idx",
  heapallindexed: true
)

// Verify indexes in the 'public' schema only
await dbContext.Database.VerifyAllIndexes(
    new VerifyAllIndexesOptions { SchemaPattern = "public" }
).ToListAsync();

// Verify indexes matching a name pattern
await dbContext.Database.VerifyAllIndexes(
    new VerifyAllIndexesOptions { IndexPattern = "search_%" }
).ToListAsync();

// Combine filters
await dbContext.Database.VerifyAllIndexes(
    new VerifyAllIndexesOptions
    {
        SchemaPattern = "public",
        IndexPattern = "search_%",
        HeapAllIndexed = true,
    }
).ToListAsync();

Listing All ParadeDB Indexes

To see all ParadeDB indexes in the database with summary statistics:

SELECT * FROM pdb.indexes();

import { diagnostics } from "@paradedb/drizzle-paradedb";

await db.execute(diagnostics.indexes());

from paradedb import paradedb_indexes

paradedb_indexes()

from paradedb.sqlalchemy import diagnostics

diagnostics.paradedb_indexes(engine)

ParadeDB.paradedb_indexes()

await dbContext.Database
    .Indexes()
    .Select(index => new
    {
        index.SchemaName,
        index.TableName,
        index.IndexName,
        index.NumSegments,
        index.TotalDocs,
    })
    .ToListAsync();

 schemaname |  tablename  |   indexname   | indexrelid | num_segments | total_docs
------------+-------------+---------------+------------+--------------+------------
 public     | products    | products_idx  |      16421 |            3 |      50000
 public     | documents   | documents_idx |      16435 |            5 |     125000
 app        | articles    | articles_idx  |      16448 |            2 |      10000

Inspecting Index Creation Time

To see when the index was built:

SQL

SELECT paradedb.index_created_at('search_idx');

       index_created_at
------------------------------
 2026-07-02 14:31:07.442218+00

Build-time stamping is supported on ParadeDB 0.24.2 and up. Indexes created before upgrading to 0.24.2 will return NULL.

Function Reference

`pdb.verify_index`

Verifies a single ParadeDB index. Parameters

Parameter	Type	Default	Description
`index`	regclass	(required)	The index to verify
`heapallindexed`	boolean	`false`	Check that all indexed ctids exist in the heap
`sample_rate`	float	`NULL`	Fraction of documents to check (0.0-1.0). NULL = check all
`report_progress`	boolean	`false`	Emit progress messages
`verbose`	boolean	`false`	Emit detailed segment-level progress and resume hints
`on_error_stop`	boolean	`false`	Stop on first error found
`segment_ids`	int[]	`NULL`	Specific segment indices to check. NULL = all segments

Return Columns

Column	Type	Description
`check_name`	text	Name of the verification check
`passed`	boolean	Whether the check passed
`details`	text	Additional information about the check result

`pdb.verify_all_indexes`

Verifies all ParadeDB indexes in the database. Parameters

Parameter	Type	Default	Description
`schema_pattern`	text	`NULL`	Filter by schema name (SQL LIKE pattern). NULL = all
`index_pattern`	text	`NULL`	Filter by index name (SQL LIKE pattern). NULL = all
`heapallindexed`	boolean	`false`	Check that all indexed ctids exist in the heap
`sample_rate`	float	`NULL`	Fraction of documents to check (0.0-1.0). NULL = check all
`report_progress`	boolean	`false`	Emit progress messages
`on_error_stop`	boolean	`false`	Stop on first error found

Return Columns

Column	Type	Description
`schemaname`	text	Schema containing the index
`indexname`	text	Name of the index
`check_name`	text	Name of the verification check
`passed`	boolean	Whether the check passed
`details`	text	Additional information about the check result

`pdb.index_segments`

Lists all segments in a ParadeDB index. Parameters

Parameter	Type	Default	Description
`index`	regclass	(required)	The index to inspect

Return Columns

Column	Type	Description
`partition_name`	text	Name of the index partition
`segment_idx`	int	Segment index (use with `segment_ids` parameter)
`segment_id`	text	Tantivy segment UUID
`num_docs`	bigint	Number of live documents
`num_deleted`	bigint	Number of deleted documents
`max_doc`	bigint	Maximum document ID

`pdb.indexes`

Lists all ParadeDB indexes in the database. Parameters None. Return Columns

Column	Type	Description
`schemaname`	text	Schema containing the index
`tablename`	text	Table the index is on
`indexname`	text	Name of the index
`indexrelid`	oid	OID of the index
`num_segments`	int	Number of Tantivy segments
`total_docs`	bigint	Total documents across all segments

`paradedb.index_created_at`

Inspects when a ParadeDB index was built. Parameters

Parameter	Type	Default	Description
`index`	regclass	(required)	The index to inspect

Return Value Returns the index creation time as a timestamp with time zone, or NULL for indexes created before upgrading to ParadeDB 0.24.2.

Deployment

Replication

Index Maintenance

Performance

Administration

Continuous Integration

Verify Index Integrity

Basic Verification

Example Output

Heap Reference Validation

Sampling for Large Indexes

Verification Settings

Progress Reporting

Verbose Mode

Stopping on First Error

Parallel Verification

Listing Segments

Verifying Specific Segments

Distributing Verification Across Workers

Verifying All ParadeDB Indexes

Filtering by Pattern

Listing All ParadeDB Indexes

Inspecting Index Creation Time

Function Reference

`pdb.verify_index`

`pdb.verify_all_indexes`

`pdb.index_segments`

`pdb.indexes`

`paradedb.index_created_at`

​Basic Verification

​Example Output

​Heap Reference Validation

​Sampling for Large Indexes

​Verification Settings

​Progress Reporting

​Verbose Mode

​Stopping on First Error

​Parallel Verification

​Listing Segments

​Verifying Specific Segments

​Distributing Verification Across Workers

​Verifying All ParadeDB Indexes

​Filtering by Pattern

​Listing All ParadeDB Indexes

​Inspecting Index Creation Time

​Function Reference

​pdb.verify_index

​pdb.verify_all_indexes

​pdb.index_segments

​pdb.indexes

​paradedb.index_created_at

Basic Verification

Example Output

Heap Reference Validation

Sampling for Large Indexes

Verification Settings

Progress Reporting

Verbose Mode

Stopping on First Error

Parallel Verification

Listing Segments

Verifying Specific Segments

Distributing Verification Across Workers

Verifying All ParadeDB Indexes

Filtering by Pattern

Listing All ParadeDB Indexes

Inspecting Index Creation Time

Function Reference

`pdb.verify_index`

`pdb.verify_all_indexes`

`pdb.index_segments`

`pdb.indexes`

`paradedb.index_created_at`