feat: expose structured Function refresh options in SDKs (#5047)

## Problem

ENT-3267 still lacks typed refresh controls for `supervisor_resources`
and `planning_workers` in Python, Rust, and Node, and Node has no
`retry_policy` refresh option. Callers must drop to SQL for those
controls.

## Change

- Add canonical Rust `FunctionRefreshPodResources` and
`FunctionRefreshPlanningConfig`, and serialize them through
`RefreshColumnOptions`.
- Expose both options in the Python sync/async wrappers and PyO3
binding.
- Expose both options in the Node NAPI object and add a Node
`retryPolicy` object backed by the existing Rust `FunctionRetryPolicy`.
- Omitted options stay omitted; branch targeting continues through the
table handle. No server, dependency, or pin changes are included.

Client-side checks only reject non-positive structural values.
Deployment resource limits and column-kind validation remain
server-side.

## Validation

`cargo fmt --all -- --check`, `python3 -m py_compile`, and `git diff
--check` pass. No native build was run locally; CI compiles and tests
the change. Generated JS API pages are unchanged.
This commit is contained in:
Xuanwo
2026-11-14 20:59:44 +08:00
committed by GitHub
parent 4843ac4c7e
commit f6f8119ba4
21 changed files with 1164 additions and 10 deletions
+8 -2
View File
@@ -1225,12 +1225,18 @@ LanceDB Enterprise it is the server's backfill job.
`REFRESH COLUMN ... WHERE ... WITH (...)` form: `where` narrows the rows,
`errorPolicy`, `maxFailedFraction`, `maxFailedRows`, `rowTimeoutSeconds`
and `initTimeoutSeconds` govern failing rows. `batchSize`, `gpuClass`,
`maxConcurrency`, `resourceBudget` and
`maxExecutorProcessesPerWorker` size registered Function refreshes.
`maxConcurrency`, `resourceBudget`, `supervisorResources`,
`planningWorkers` and `maxExecutorProcessesPerWorker` size registered
Function refreshes. `retryPolicy` overrides user-code retries only when
repeating partial work is safe; set `maxAttempts: 1` to disable retries.
`maxConcurrency` overrides the registered default for this logical
execution only and is a positive integer; it is not an account-wide
quota. A `resourceBudget` requires `cpuMillis` and
`memoryBytes`; its `gpuCount` defaults to zero for a CPU-only Function.
`supervisorResources` and each `planningWorkers.resources` entry require
positive `cpuMillis`, `memoryBytes` and `ephemeralStorageBytes`.
`planningWorkers` also requires positive `maxWorkers`,
`rowsPerWorker` and `fragmentsPerWorker`.
The budget applies only to this explicit refresh and does not configure
a later automatic refresh.
`gpuCount` at the top level is the shorthand that preserves the
+3
View File
@@ -103,6 +103,9 @@
- [FunctionErrorRecord](interfaces/FunctionErrorRecord.md)
- [FunctionErrors](interfaces/FunctionErrors.md)
- [FunctionErrorsOptions](interfaces/FunctionErrorsOptions.md)
- [FunctionRefreshPlanningConfig](interfaces/FunctionRefreshPlanningConfig.md)
- [FunctionRefreshPodResources](interfaces/FunctionRefreshPodResources.md)
- [FunctionRetryPolicy](interfaces/FunctionRetryPolicy.md)
- [FunctionWorkerResourceBudget](interfaces/FunctionWorkerResourceBudget.md)
- [HnswPqOptions](interfaces/HnswPqOptions.md)
- [HnswSqOptions](interfaces/HnswSqOptions.md)
@@ -0,0 +1,50 @@
[**@lancedb/lancedb**](../README.md) • **Docs**
***
[@lancedb/lancedb](../globals.md) / FunctionRefreshPlanningConfig
# Interface: FunctionRefreshPlanningConfig
Resources and sizing policy for the planning fleet of one registered
Function refresh.
## Properties
### fragmentsPerWorker
```ts
fragmentsPerWorker: number;
```
Maximum fragments assigned to one planning worker.
***
### maxWorkers
```ts
maxWorkers: number;
```
Maximum number of planning workers.
***
### resources
```ts
resources: FunctionRefreshPodResources;
```
CPU, memory, and ephemeral-storage limits for one planning worker.
***
### rowsPerWorker
```ts
rowsPerWorker: number;
```
Maximum rows assigned to one planning worker.
@@ -0,0 +1,40 @@
[**@lancedb/lancedb**](../README.md) • **Docs**
***
[@lancedb/lancedb](../globals.md) / FunctionRefreshPodResources
# Interface: FunctionRefreshPodResources
CPU, memory, and ephemeral-storage limits for one Function refresh
control-plane pod.
## Properties
### cpuMillis
```ts
cpuMillis: number;
```
CPU limit in millicores; 1000 is one CPU.
***
### ephemeralStorageBytes
```ts
ephemeralStorageBytes: number;
```
Ephemeral storage limit in bytes.
***
### memoryBytes
```ts
memoryBytes: number;
```
Memory limit in bytes.
@@ -0,0 +1,91 @@
[**@lancedb/lancedb**](../README.md) • **Docs**
***
[@lancedb/lancedb](../globals.md) / FunctionRetryPolicy
# Interface: FunctionRetryPolicy
Overrides retries of failed user-code calls for one refresh. Omitted
fields use the Function refresh defaults.
## Properties
### errorCodes?
```ts
optional errorCodes: string[];
```
Structured provider error codes or HTTP status codes.
***
### exceptionTypes?
```ts
optional exceptionTypes: string[];
```
Exception class names, preferably qualified.
***
### honorRetryAfter?
```ts
optional honorRetryAfter: boolean;
```
Honor Retry-After within the configured maximum.
***
### initialBackoffSeconds?
```ts
optional initialBackoffSeconds: number;
```
Initial exponential-backoff ceiling, in seconds.
***
### maxAttempts?
```ts
optional maxAttempts: number;
```
Total calls, including the first; 1 disables retries, maximum 100.
***
### maxBackoffSeconds?
```ts
optional maxBackoffSeconds: number;
```
Maximum backoff, in seconds.
***
### messagePattern?
```ts
optional messagePattern: string;
```
Regular expression matched against the exception class and message.
***
### transient?
```ts
optional transient: boolean;
```
Match connection/timeouts, HTTP 429/503, SlowDown and
ServiceUnavailable.
@@ -117,6 +117,16 @@ Worker count for an expression column; Function columns reject it.
***
### planningWorkers?
```ts
optional planningWorkers: FunctionRefreshPlanningConfig;
```
Resources and sizing for the refresh planning fleet.
***
### publicationIntervalRows?
```ts
@@ -141,6 +151,16 @@ refresh. `gpuCount` inside the budget defaults to zero.
***
### retryPolicy?
```ts
optional retryPolicy: FunctionRetryPolicy;
```
Override retries of failed user-code calls for this refresh.
***
### rowTimeoutSeconds?
```ts
@@ -151,6 +171,16 @@ Deadline for one Function call, in seconds.
***
### supervisorResources?
```ts
optional supervisorResources: FunctionRefreshPodResources;
```
CPU, memory, and ephemeral-storage limits for the refresh supervisor.
***
### where?
```ts
@@ -400,6 +400,101 @@ for changing only the aggregate GPU budget. It cannot be combined with the
complete resource budget. Omitting both preserves the server's existing
defaults.
{/*
Draft for the Function refresh control-plane resource release.
`resource_budget` above sizes the Function workers. Two independent options
size the refresh control plane for a registered Function.
`supervisor_resources` sets the CPU, memory and ephemeral-storage limits of
the supervisor pod. `planning_workers` sets the resources of one planning
worker and the sizing of the CPU-only planning fleet: `max_workers`,
`rows_per_worker` and `fragments_per_worker`. Every field is required and
must be positive. Omitting either option leaves the server's configured
defaults unchanged, and either can be combined with `resource_budget`.
<CodeGroup>
```python Python icon="python"
job = table.refresh_column_async(
"embedding",
supervisor_resources={
"cpu_millis": 2_000,
"memory_bytes": 4 * 1024**3,
"ephemeral_storage_bytes": 10 * 1024**3,
},
planning_workers={
"resources": {
"cpu_millis": 1_000,
"memory_bytes": 2 * 1024**3,
"ephemeral_storage_bytes": 5 * 1024**3,
},
"max_workers": 4,
"rows_per_worker": 100_000,
"fragments_per_worker": 8,
},
)
job.wait()
```
```rust Rust icon="rust"
use lancedb::table::{
FunctionRefreshPlanningConfig,
FunctionRefreshPodResources,
RefreshColumnOptions,
};
let _job = table
.refresh_column_async_with_options(
"embedding",
RefreshColumnOptions::default()
.supervisor_resources(
FunctionRefreshPodResources::new(
2_000,
4 * 1024 * 1024 * 1024,
10 * 1024 * 1024 * 1024,
),
)
.planning_workers(
FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(
1_000,
2 * 1024 * 1024 * 1024,
5 * 1024 * 1024 * 1024,
),
4,
100_000,
8,
),
),
)
.await?;
```
```typescript TypeScript icon="typescript"
const job = await table.refreshColumnAsync("embedding", {
supervisorResources: {
cpuMillis: 2_000,
memoryBytes: 4 * 1024 ** 3,
ephemeralStorageBytes: 10 * 1024 ** 3,
},
planningWorkers: {
resources: {
cpuMillis: 1_000,
memoryBytes: 2 * 1024 ** 3,
ephemeralStorageBytes: 5 * 1024 ** 3,
},
maxWorkers: 4,
rowsPerWorker: 100_000,
fragmentsPerWorker: 8,
},
});
await job.wait();
```
</CodeGroup>
The SDK validates the shape and sends these values only with this explicit
refresh request. A supporting LanceDB Enterprise server is required.
*/}
These examples configure only the explicit refresh shown. A registered
Function declaration can separately set `auto_refresh_resource_budget` on
`add_columns`; that declaration controls the budget for its automatic
@@ -457,6 +552,39 @@ from `lancedb.functions`. Use `FunctionRetryPolicy(max_attempts=1)` to
disable user-error retries. Omitting the override preserves the registered
policy.
<CodeGroup>
```python Python icon="python"
from lancedb.functions import FunctionRetryPolicy
job = table.refresh_column_async(
"embedding",
retry_policy=FunctionRetryPolicy(transient=True, max_attempts=3),
)
job.wait()
```
```rust Rust icon="rust"
use lancedb::{function::FunctionRetryPolicy, table::RefreshColumnOptions};
let mut retry_policy = FunctionRetryPolicy::transient();
retry_policy.max_attempts = 3;
let _job = table
.refresh_column_async_with_options(
"embedding",
RefreshColumnOptions::default().retry_policy(retry_policy),
)
.await?;
```
```typescript TypeScript icon="typescript"
const job = await table.refreshColumnAsync("embedding", {
retryPolicy: { transient: true, maxAttempts: 3 },
});
await job.wait();
```
</CodeGroup>
This requires a supporting LanceDB Enterprise server as well as an SDK
that exposes these options. An older Function image may need
re-registration before it accepts a refresh override. Only enable retries
+121
View File
@@ -979,6 +979,42 @@ describe("remote connection", () => {
},
"resourceBudget.gpuCount",
],
[
{
supervisorResources: {
cpuMillis: 0,
memoryBytes: 1_024,
ephemeralStorageBytes: 2_048,
},
},
"supervisorResources.cpuMillis",
],
[
{
supervisorResources: {
cpuMillis: 1_000,
memoryBytes: 1_024,
ephemeralStorageBytes: 0,
},
},
"supervisorResources.ephemeralStorageBytes",
],
[
{
planningWorkers: {
resources: {
cpuMillis: 1_000,
memoryBytes: 1_024,
ephemeralStorageBytes: 2_048,
},
maxWorkers: 0,
rowsPerWorker: 100,
fragmentsPerWorker: 1,
},
},
"planningWorkers.maxWorkers",
],
[{ retryPolicy: { maxAttempts: 0 } }, "retry_policy.max_attempts"],
] as const) {
await expect(
table.refreshColumnAsync("embedding", options),
@@ -1016,6 +1052,91 @@ describe("remote connection", () => {
]);
});
it("forwards structured Function refresh options and Node retry policy", async () => {
const bodies: unknown[] = [];
await withMockDatabase(
(req, res) => {
const path = req.url ?? "";
if (path.endsWith("/describe/")) {
res.writeHead(200, { "Content-Type": "application/json" }).end(
JSON.stringify({
name: "docs",
version: 1,
schema: { fields: [] },
}),
);
return;
}
if (path === "/v1/table/docs/backfill_column") {
let body = "";
req.on("data", (chunk) => {
body += chunk;
});
req.on("end", () => {
bodies.push(JSON.parse(body));
res
.writeHead(202, { "Content-Type": "application/json" })
.end('{"job_id": "j-structured"}');
});
return;
}
res.writeHead(404).end();
},
async (db) => {
const table = await db.openTable("docs");
const job = await table.refreshColumnAsync("embedding", {
supervisorResources: {
cpuMillis: 2_000,
memoryBytes: 4 * 1024 ** 3,
ephemeralStorageBytes: 10 * 1024 ** 3,
},
planningWorkers: {
resources: {
cpuMillis: 1_000,
memoryBytes: 2 * 1024 ** 3,
ephemeralStorageBytes: 5 * 1024 ** 3,
},
maxWorkers: 4,
rowsPerWorker: 100_000,
fragmentsPerWorker: 8,
},
retryPolicy: { transient: true },
});
expect(job.id).toBe("j-structured");
},
);
expect(bodies).toEqual([
{
column: "embedding",
["supervisor_resources"]: {
["cpu_millis"]: 2_000,
["memory_bytes"]: 4 * 1024 ** 3,
["ephemeral_storage_bytes"]: 10 * 1024 ** 3,
},
["planning_workers"]: {
resources: {
["cpu_millis"]: 1_000,
["memory_bytes"]: 2 * 1024 ** 3,
["ephemeral_storage_bytes"]: 5 * 1024 ** 3,
},
["max_workers"]: 4,
["rows_per_worker"]: 100_000,
["fragments_per_worker"]: 8,
},
["retry_policy"]: {
["max_attempts"]: 3,
["initial_backoff_seconds"]: 0.2,
["max_backoff_seconds"]: 30,
["exception_types"]: [],
["message_pattern"]: null,
["error_codes"]: [],
transient: true,
["honor_retry_after"]: true,
},
},
]);
});
it("surfaces JSON server errors from remote table operations", async () => {
await withMockDatabase(
(req, res) => {
+25
View File
@@ -5716,6 +5716,31 @@ describe("computed columns", () => {
resourceBudget: { cpuMillis: 1_000, memoryBytes: 1_024 },
}),
).rejects.toThrow("resource_budget requires a LanceDB Enterprise refresh");
await expect(
table.refreshColumn("x", {
supervisorResources: {
cpuMillis: 1_000,
memoryBytes: 1_024,
ephemeralStorageBytes: 2_048,
},
}),
).rejects.toThrow(
"supervisor_resources requires a LanceDB Enterprise refresh",
);
await expect(
table.refreshColumnAsync("x", {
planningWorkers: {
resources: {
cpuMillis: 1_000,
memoryBytes: 1_024,
ephemeralStorageBytes: 2_048,
},
maxWorkers: 2,
rowsPerWorker: 100,
fragmentsPerWorker: 1,
},
}),
).rejects.toThrow("planning_workers requires a LanceDB Enterprise refresh");
});
it("records Function errors only on remote tables", async () => {
+3
View File
@@ -57,6 +57,9 @@ export {
AddResult,
AddColumnsResult,
RefreshColumnOptions,
FunctionRefreshPlanningConfig,
FunctionRefreshPodResources,
FunctionRetryPolicy,
FunctionWorkerResourceBudget,
RefreshColumnResult,
FunctionErrors,
+8 -2
View File
@@ -831,12 +831,18 @@ export abstract class Table {
* `REFRESH COLUMN ... WHERE ... WITH (...)` form: `where` narrows the rows,
* `errorPolicy`, `maxFailedFraction`, `maxFailedRows`, `rowTimeoutSeconds`
* and `initTimeoutSeconds` govern failing rows. `batchSize`, `gpuClass`,
* `maxConcurrency`, `resourceBudget` and
* `maxExecutorProcessesPerWorker` size registered Function refreshes.
* `maxConcurrency`, `resourceBudget`, `supervisorResources`,
* `planningWorkers` and `maxExecutorProcessesPerWorker` size registered
* Function refreshes. `retryPolicy` overrides user-code retries only when
* repeating partial work is safe; set `maxAttempts: 1` to disable retries.
* `maxConcurrency` overrides the registered default for this logical
* execution only and is a positive integer; it is not an account-wide
* quota. A `resourceBudget` requires `cpuMillis` and
* `memoryBytes`; its `gpuCount` defaults to zero for a CPU-only Function.
* `supervisorResources` and each `planningWorkers.resources` entry require
* positive `cpuMillis`, `memoryBytes` and `ephemeralStorageBytes`.
* `planningWorkers` also requires positive `maxWorkers`,
* `rowsPerWorker` and `fragmentsPerWorker`.
* The budget applies only to this explicit refresh and does not configure
* a later automatic refresh.
* `gpuCount` at the top level is the shorthand that preserves the
+140
View File
@@ -9,6 +9,8 @@ use lancedb::ipc::{ipc_file_to_batches, ipc_file_to_schema};
use lancedb::table::{
AddDataMode, ColumnAlteration as LanceColumnAlteration,
FieldMetadataUpdate as LanceFieldMetadataUpdate, FtsToken as LanceDbFtsToken,
FunctionRefreshPlanningConfig as LanceFunctionRefreshPlanningConfig,
FunctionRefreshPodResources as LanceFunctionRefreshPodResources,
FunctionWorkerResourceBudget as LanceFunctionWorkerResourceBudget, NewColumnTransform, Ref,
Table as LanceDbTable,
};
@@ -1705,6 +1707,58 @@ pub struct FunctionWorkerResourceBudget {
pub gpu_count: Option<f64>,
}
/// CPU, memory, and ephemeral-storage limits for one Function refresh
/// control-plane pod.
#[napi(object)]
#[derive(Clone, Debug)]
pub struct FunctionRefreshPodResources {
/// CPU limit in millicores; 1000 is one CPU.
pub cpu_millis: f64,
/// Memory limit in bytes.
pub memory_bytes: f64,
/// Ephemeral storage limit in bytes.
pub ephemeral_storage_bytes: f64,
}
/// Resources and sizing policy for the planning fleet of one registered
/// Function refresh.
#[napi(object)]
#[derive(Clone, Debug)]
pub struct FunctionRefreshPlanningConfig {
/// CPU, memory, and ephemeral-storage limits for one planning worker.
pub resources: FunctionRefreshPodResources,
/// Maximum number of planning workers.
pub max_workers: f64,
/// Maximum rows assigned to one planning worker.
pub rows_per_worker: f64,
/// Maximum fragments assigned to one planning worker.
pub fragments_per_worker: f64,
}
/// Overrides retries of failed user-code calls for one refresh. Omitted
/// fields use the Function refresh defaults.
#[napi(object)]
#[derive(Clone, Debug, Default)]
pub struct FunctionRetryPolicy {
/// Total calls, including the first; 1 disables retries, maximum 100.
pub max_attempts: Option<u32>,
/// Initial exponential-backoff ceiling, in seconds.
pub initial_backoff_seconds: Option<f64>,
/// Maximum backoff, in seconds.
pub max_backoff_seconds: Option<f64>,
/// Exception class names, preferably qualified.
pub exception_types: Option<Vec<String>>,
/// Regular expression matched against the exception class and message.
pub message_pattern: Option<String>,
/// Structured provider error codes or HTTP status codes.
pub error_codes: Option<Vec<String>>,
/// Match connection/timeouts, HTTP 429/503, SlowDown and
/// ServiceUnavailable.
pub transient: Option<bool>,
/// Honor Retry-After within the configured maximum.
pub honor_retry_after: Option<bool>,
}
fn positive_safe_integer(value: f64, name: &str) -> napi::Result<u64> {
const MAX_SAFE_INTEGER: f64 = 9_007_199_254_740_991.0;
if !value.is_finite() || value.fract() != 0.0 || !(1.0..=MAX_SAFE_INTEGER).contains(&value) {
@@ -1768,8 +1822,14 @@ pub struct RefreshColumnOptions {
/// Complete aggregate CPU and memory budget for a registered Function
/// refresh. `gpuCount` inside the budget defaults to zero.
pub resource_budget: Option<FunctionWorkerResourceBudget>,
/// CPU, memory, and ephemeral-storage limits for the refresh supervisor.
pub supervisor_resources: Option<FunctionRefreshPodResources>,
/// Resources and sizing for the refresh planning fleet.
pub planning_workers: Option<FunctionRefreshPlanningConfig>,
/// Positive aggregate GPU budget for Function workers.
pub gpu_count: Option<u32>,
/// Override retries of failed user-code calls for this refresh.
pub retry_policy: Option<FunctionRetryPolicy>,
/// `"fail"` (default) stops on a row error; `"skip"` writes null and
/// records the error, subject to the failure limits.
pub error_policy: Option<String>,
@@ -1814,7 +1874,87 @@ impl TryFrom<RefreshColumnOptions> for lancedb::table::RefreshColumnOptions {
)
})
.transpose()?;
options.supervisor_resources = value
.supervisor_resources
.map(|resources| {
Ok::<_, napi::Error>(LanceFunctionRefreshPodResources::new(
positive_safe_integer(resources.cpu_millis, "supervisorResources.cpuMillis")?,
positive_safe_integer(
resources.memory_bytes,
"supervisorResources.memoryBytes",
)?,
positive_safe_integer(
resources.ephemeral_storage_bytes,
"supervisorResources.ephemeralStorageBytes",
)?,
))
})
.transpose()?;
options.planning_workers = value
.planning_workers
.map(|planning| {
let resources = LanceFunctionRefreshPodResources::new(
positive_safe_integer(
planning.resources.cpu_millis,
"planningWorkers.resources.cpuMillis",
)?,
positive_safe_integer(
planning.resources.memory_bytes,
"planningWorkers.resources.memoryBytes",
)?,
positive_safe_integer(
planning.resources.ephemeral_storage_bytes,
"planningWorkers.resources.ephemeralStorageBytes",
)?,
);
let max_workers = usize::try_from(positive_safe_integer(
planning.max_workers,
"planningWorkers.maxWorkers",
)?)
.map_err(|_| napi::Error::from_reason("planningWorkers.maxWorkers is too large"))?;
Ok::<_, napi::Error>(LanceFunctionRefreshPlanningConfig::new(
resources,
max_workers,
positive_safe_integer(
planning.rows_per_worker,
"planningWorkers.rowsPerWorker",
)?,
positive_safe_integer(
planning.fragments_per_worker,
"planningWorkers.fragmentsPerWorker",
)?,
))
})
.transpose()?;
options.gpu_count = value.gpu_count;
options.retry_policy = value.retry_policy.map(|policy| {
let mut converted = lancedb::function::FunctionRetryPolicy::default();
if let Some(max_attempts) = policy.max_attempts {
converted.max_attempts = max_attempts;
}
if let Some(seconds) = policy.initial_backoff_seconds {
converted.initial_backoff_seconds = seconds;
}
if let Some(seconds) = policy.max_backoff_seconds {
converted.max_backoff_seconds = seconds;
}
if let Some(exception_types) = policy.exception_types {
converted.exception_types = exception_types;
}
if let Some(pattern) = policy.message_pattern {
converted.message_pattern = Some(pattern);
}
if let Some(error_codes) = policy.error_codes {
converted.error_codes = error_codes;
}
if let Some(transient) = policy.transient {
converted.transient = transient;
}
if let Some(honor_retry_after) = policy.honor_retry_after {
converted.honor_retry_after = honor_retry_after;
}
converted
});
options.error_policy = match value.error_policy.as_deref() {
None => None,
Some("fail") => Some(lancedb::function::FunctionErrorPolicy::Fail),
+4
View File
@@ -663,6 +663,8 @@ class Table:
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[dict[str, int]] = None,
supervisor_resources: Optional[dict[str, int]] = None,
planning_workers: Optional[dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[str] = None,
@@ -683,6 +685,8 @@ class Table:
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[dict[str, int]] = None,
supervisor_resources: Optional[dict[str, int]] = None,
planning_workers: Optional[dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[str] = None,
+8
View File
@@ -1164,6 +1164,8 @@ class RemoteTable(Table):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -1183,6 +1185,8 @@ class RemoteTable(Table):
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=retry_policy,
@@ -1205,6 +1209,8 @@ class RemoteTable(Table):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -1225,6 +1231,8 @@ class RemoteTable(Table):
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=retry_policy,
+64
View File
@@ -3258,6 +3258,8 @@ class Table(ABC):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -3315,6 +3317,17 @@ class Table(ABC):
CPU-only budget. This budget cannot be combined with the top-level
``gpu_count`` shorthand. LanceDB Enterprise only; local refreshes
reject it.
supervisor_resources: dict[str, int], optional
CPU, memory, and ephemeral-storage limits for the Function refresh
supervisor pod. Keys are ``cpu_millis``, ``memory_bytes`` and
``ephemeral_storage_bytes``; all are required and positive.
LanceDB Enterprise only; local refreshes reject it.
planning_workers: dict[str, object], optional
Resources and sizing for the Function refresh planning fleet.
``resources`` has the same keys as ``supervisor_resources``; the
other keys are ``max_workers``, ``rows_per_worker`` and
``fragments_per_worker``. LanceDB Enterprise only; local refreshes
reject it.
gpu_class: str, optional
GPU class for Function workers. Must be a lowercase name of 1-63
letters, digits, or hyphens, without a leading or trailing hyphen.
@@ -3367,6 +3380,8 @@ class Table(ABC):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -3423,6 +3438,17 @@ class Table(ABC):
CPU-only budget. This budget cannot be combined with the top-level
``gpu_count`` shorthand. LanceDB Enterprise only; local refreshes
reject it.
supervisor_resources: dict[str, int], optional
CPU, memory, and ephemeral-storage limits for the Function refresh
supervisor pod. Keys are ``cpu_millis``, ``memory_bytes`` and
``ephemeral_storage_bytes``; all are required and positive.
LanceDB Enterprise only; local refreshes reject it.
planning_workers: dict[str, object], optional
Resources and sizing for the Function refresh planning fleet.
``resources`` has the same keys as ``supervisor_resources``; the
other keys are ``max_workers``, ``rows_per_worker`` and
``fragments_per_worker``. LanceDB Enterprise only; local refreshes
reject it.
gpu_class: str, optional
GPU class for Function workers. Must be a lowercase name of 1-63
letters, digits, or hyphens, without a leading or trailing hyphen.
@@ -5632,6 +5658,8 @@ class LanceTable(Table):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -5654,6 +5682,8 @@ class LanceTable(Table):
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=retry_policy,
@@ -5676,6 +5706,8 @@ class LanceTable(Table):
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -5700,6 +5732,8 @@ class LanceTable(Table):
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=retry_policy,
@@ -8109,6 +8143,8 @@ class AsyncTable:
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -8166,6 +8202,17 @@ class AsyncTable:
CPU-only budget. This budget cannot be combined with the top-level
``gpu_count`` shorthand. LanceDB Enterprise only; local refreshes
reject it.
supervisor_resources: dict[str, int], optional
CPU, memory, and ephemeral-storage limits for the Function refresh
supervisor pod. Keys are ``cpu_millis``, ``memory_bytes`` and
``ephemeral_storage_bytes``; all are required and positive.
LanceDB Enterprise only; local refreshes reject it.
planning_workers: dict[str, object], optional
Resources and sizing for the Function refresh planning fleet.
``resources`` has the same keys as ``supervisor_resources``; the
other keys are ``max_workers``, ``rows_per_worker`` and
``fragments_per_worker``. LanceDB Enterprise only; local refreshes
reject it.
gpu_class: str, optional
GPU class for Function workers. Must be a lowercase name of 1-63
letters, digits, or hyphens, without a leading or trailing hyphen.
@@ -8213,6 +8260,8 @@ class AsyncTable:
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=None
@@ -8351,6 +8400,8 @@ class AsyncTable:
max_concurrency: Optional[int] = None,
publication_interval_rows: Optional[int] = None,
resource_budget: Optional[Dict[str, int]] = None,
supervisor_resources: Optional[Dict[str, int]] = None,
planning_workers: Optional[Dict[str, Any]] = None,
gpu_class: Optional[str] = None,
gpu_count: Optional[int] = None,
retry_policy: Optional[FunctionRetryPolicy] = None,
@@ -8407,6 +8458,17 @@ class AsyncTable:
CPU-only budget. This budget cannot be combined with the top-level
``gpu_count`` shorthand. LanceDB Enterprise only; local refreshes
reject it.
supervisor_resources: dict[str, int], optional
CPU, memory, and ephemeral-storage limits for the Function refresh
supervisor pod. Keys are ``cpu_millis``, ``memory_bytes`` and
``ephemeral_storage_bytes``; all are required and positive.
LanceDB Enterprise only; local refreshes reject it.
planning_workers: dict[str, object], optional
Resources and sizing for the Function refresh planning fleet.
``resources`` has the same keys as ``supervisor_resources``; the
other keys are ``max_workers``, ``rows_per_worker`` and
``fragments_per_worker``. LanceDB Enterprise only; local refreshes
reject it.
gpu_class: str, optional
GPU class for Function workers. Must be a lowercase name of 1-63
letters, digits, or hyphens, without a leading or trailing hyphen.
@@ -8471,6 +8533,8 @@ class AsyncTable:
max_concurrency=max_concurrency,
publication_interval_rows=publication_interval_rows,
resource_budget=resource_budget,
supervisor_resources=supervisor_resources,
planning_workers=planning_workers,
gpu_class=gpu_class,
gpu_count=gpu_count,
retry_policy=None
+65
View File
@@ -2076,6 +2076,23 @@ def test_remote_create_index_async_returns_job():
"gpu_count": 2,
}
},
{
"supervisor_resources": {
"cpu_millis": 2000,
"memory_bytes": 4 * 1024**3,
"ephemeral_storage_bytes": 10 * 1024**3,
},
"planning_workers": {
"resources": {
"cpu_millis": 1000,
"memory_bytes": 2 * 1024**3,
"ephemeral_storage_bytes": 5 * 1024**3,
},
"max_workers": 4,
"rows_per_worker": 100_000,
"fragments_per_worker": 8,
},
},
{"gpu_class": "a10g", "gpu_count": 2},
{
"retry_policy": {"transient": True, "max_attempts": 3},
@@ -2122,6 +2139,9 @@ async def test_remote_refresh_async_returns_typed_terminal_result(
**budget,
"gpu_count": budget.get("gpu_count", 0),
}
for key in ("supervisor_resources", "planning_workers"):
if refresh_kwargs.get(key) is not None:
expected[key] = refresh_kwargs[key]
if refresh_kwargs.get("retry_policy") is not None:
expected["retry_policy"] = lancedb.FunctionRetryPolicy.model_validate(
refresh_kwargs["retry_policy"]
@@ -2350,6 +2370,51 @@ def test_remote_refresh_async_forwards_progress_from_submitted_handle():
},
"unsupported key 'unknown'",
),
(
{
"supervisor_resources": {
"cpu_millis": 0,
"memory_bytes": 1,
"ephemeral_storage_bytes": 1,
}
},
"supervisor_resources.cpu_millis",
),
(
{
"supervisor_resources": {
"cpu_millis": 1,
"memory_bytes": 1,
}
},
"supervisor_resources.ephemeral_storage_bytes",
),
(
{
"supervisor_resources": {
"cpu_millis": 1,
"memory_bytes": 1,
"ephemeral_storage_bytes": 1,
"unknown": 1,
}
},
"unsupported key 'unknown'",
),
(
{
"planning_workers": {
"resources": {
"cpu_millis": 1,
"memory_bytes": 1,
"ephemeral_storage_bytes": 1,
},
"max_workers": 0,
"rows_per_worker": 1,
"fragments_per_worker": 1,
}
},
"planning_workers.max_workers",
),
(
{"max_executor_processes_per_worker": -2},
"max_executor_processes_per_worker",
+25
View File
@@ -6013,6 +6013,31 @@ def test_refresh_column_rejects_invalid_batch_size(tmp_path, method, batch_size)
{"resource_budget": {"cpu_millis": 1000, "memory_bytes": 1024}},
"resource_budget",
),
(
{
"supervisor_resources": {
"cpu_millis": 1000,
"memory_bytes": 1024,
"ephemeral_storage_bytes": 2048,
}
},
"supervisor_resources",
),
(
{
"planning_workers": {
"resources": {
"cpu_millis": 1000,
"memory_bytes": 1024,
"ephemeral_storage_bytes": 2048,
},
"max_workers": 2,
"rows_per_worker": 100,
"fragments_per_worker": 1,
}
},
"planning_workers",
),
({"num_workers": 2}, "num_workers"),
({"max_executor_processes_per_worker": 2}, "max_executor_processes_per_worker"),
],
+114 -4
View File
@@ -30,8 +30,9 @@ use lancedb::connection::LanceFileVersion;
use lancedb::index::scalar::FtsIndexBuilder;
use lancedb::table::{
AddDataMode, ColumnAlteration, CompactionMode, CompactionOptions, Duration,
FieldMetadataUpdate, FtsToken as LanceDbFtsToken, FunctionWorkerResourceBudget, IndexRemapMode,
NewColumnTransform, Ref, RefreshColumnOptions, Table as LanceDbTable,
FieldMetadataUpdate, FtsToken as LanceDbFtsToken, FunctionRefreshPlanningConfig,
FunctionRefreshPodResources, FunctionWorkerResourceBudget, IndexRemapMode, NewColumnTransform,
Ref, RefreshColumnOptions, Table as LanceDbTable,
};
use lancedb::tokenize as lancedb_tokenize;
use pyo3::{
@@ -52,6 +53,8 @@ struct RefreshColumnArgs {
max_concurrency: Option<u32>,
publication_interval_rows: Option<i64>,
resource_budget: Option<FunctionWorkerResourceBudget>,
supervisor_resources: Option<FunctionRefreshPodResources>,
planning_workers: Option<FunctionRefreshPlanningConfig>,
gpu_class: Option<String>,
gpu_count: Option<i64>,
retry_policy: Option<String>,
@@ -159,6 +162,101 @@ fn parse_resource_budget(
))
}
fn structured_integer(value: &Bound<'_, PyAny>, name: &str) -> PyResult<u64> {
if value.is_none() || value.is_instance_of::<PyBool>() {
return Err(PyValueError::new_err(format!("{name} must be an integer")));
}
let parsed = value
.extract::<u64>()
.map_err(|_| PyValueError::new_err(format!("{name} must be an integer")))?;
if parsed == 0 {
return Err(PyValueError::new_err(format!(
"{name} must be greater than zero"
)));
}
Ok(parsed)
}
fn parse_pod_resources(
resources: &Bound<'_, PyDict>,
prefix: &str,
) -> PyResult<FunctionRefreshPodResources> {
for key in resources.keys() {
let key = key
.extract::<String>()
.map_err(|_| PyValueError::new_err(format!("{prefix} keys must be strings")))?;
if !matches!(
key.as_str(),
"cpu_millis" | "memory_bytes" | "ephemeral_storage_bytes"
) {
return Err(PyValueError::new_err(format!(
"{prefix} contains unsupported key '{key}'"
)));
}
}
let get = |name: &str| -> PyResult<u64> {
let value = resources
.get_item(name)?
.ok_or_else(|| PyValueError::new_err(format!("{prefix}.{name} is required")))?;
structured_integer(&value, &format!("{prefix}.{name}"))
};
Ok(FunctionRefreshPodResources::new(
get("cpu_millis")?,
get("memory_bytes")?,
get("ephemeral_storage_bytes")?,
))
}
fn parse_supervisor_resources(
resources: Option<&Bound<'_, PyDict>>,
) -> PyResult<Option<FunctionRefreshPodResources>> {
resources
.map(|resources| parse_pod_resources(resources, "supervisor_resources"))
.transpose()
}
fn parse_planning_workers(
planning: Option<&Bound<'_, PyDict>>,
) -> PyResult<Option<FunctionRefreshPlanningConfig>> {
let Some(planning) = planning else {
return Ok(None);
};
for key in planning.keys() {
let key = key
.extract::<String>()
.map_err(|_| PyValueError::new_err("planning_workers keys must be strings"))?;
if !matches!(
key.as_str(),
"resources" | "max_workers" | "rows_per_worker" | "fragments_per_worker"
) {
return Err(PyValueError::new_err(format!(
"planning_workers contains unsupported key '{key}'"
)));
}
}
let resources = planning
.get_item("resources")?
.ok_or_else(|| PyValueError::new_err("planning_workers.resources is required"))?;
let resources = resources
.cast::<PyDict>()
.map_err(|_| PyValueError::new_err("planning_workers.resources must be a dict"))?;
let resources = parse_pod_resources(resources, "planning_workers.resources")?;
let get = |name: &str| -> PyResult<u64> {
let value = planning
.get_item(name)?
.ok_or_else(|| PyValueError::new_err(format!("planning_workers.{name} is required")))?;
structured_integer(&value, &format!("planning_workers.{name}"))
};
let max_workers = usize::try_from(get("max_workers")?)
.map_err(|_| PyValueError::new_err("planning_workers.max_workers is too large"))?;
Ok(Some(FunctionRefreshPlanningConfig::new(
resources,
max_workers,
get("rows_per_worker")?,
get("fragments_per_worker")?,
)))
}
fn refresh_column_options(args: RefreshColumnArgs) -> PyResult<RefreshColumnOptions> {
let RefreshColumnArgs {
where_clause,
@@ -166,6 +264,8 @@ fn refresh_column_options(args: RefreshColumnArgs) -> PyResult<RefreshColumnOpti
max_concurrency,
publication_interval_rows,
resource_budget,
supervisor_resources,
planning_workers,
gpu_class,
gpu_count,
retry_policy,
@@ -211,6 +311,8 @@ fn refresh_column_options(args: RefreshColumnArgs) -> PyResult<RefreshColumnOpti
options.publication_interval_rows =
positive_count(publication_interval_rows, "publication_interval_rows")?;
options.resource_budget = resource_budget;
options.supervisor_resources = supervisor_resources;
options.planning_workers = planning_workers;
options.gpu_class = gpu_class;
options.gpu_count = gpu_count
.map(|count| {
@@ -2492,7 +2594,7 @@ impl Table {
})
}
#[pyo3(signature = (column, *, r#where=None, batch_size=None, max_concurrency=None, publication_interval_rows=None, resource_budget=None, gpu_class=None, gpu_count=None, retry_policy=None, error_policy=None, max_failed_fraction=None, max_failed_rows=None, row_timeout_seconds=None, init_timeout_seconds=None, num_workers=None, max_executor_processes_per_worker=None))]
#[pyo3(signature = (column, *, r#where=None, batch_size=None, max_concurrency=None, publication_interval_rows=None, resource_budget=None, supervisor_resources=None, planning_workers=None, gpu_class=None, gpu_count=None, retry_policy=None, error_policy=None, max_failed_fraction=None, max_failed_rows=None, row_timeout_seconds=None, init_timeout_seconds=None, num_workers=None, max_executor_processes_per_worker=None))]
#[allow(clippy::too_many_arguments)]
pub fn refresh_column<'py>(
self_: PyRef<'py, Self>,
@@ -2502,6 +2604,8 @@ impl Table {
max_concurrency: Option<Bound<'py, PyAny>>,
publication_interval_rows: Option<i64>,
resource_budget: Option<Bound<'py, PyDict>>,
supervisor_resources: Option<Bound<'py, PyDict>>,
planning_workers: Option<Bound<'py, PyDict>>,
gpu_class: Option<String>,
gpu_count: Option<i64>,
retry_policy: Option<String>,
@@ -2519,6 +2623,8 @@ impl Table {
max_concurrency: positive_u32_integer(max_concurrency.as_ref(), "max_concurrency")?,
publication_interval_rows,
resource_budget: parse_resource_budget(resource_budget.as_ref())?,
supervisor_resources: parse_supervisor_resources(supervisor_resources.as_ref())?,
planning_workers: parse_planning_workers(planning_workers.as_ref())?,
gpu_class,
gpu_count,
retry_policy,
@@ -2540,7 +2646,7 @@ impl Table {
})
}
#[pyo3(signature = (column, *, r#where=None, batch_size=None, max_concurrency=None, publication_interval_rows=None, resource_budget=None, gpu_class=None, gpu_count=None, retry_policy=None, error_policy=None, max_failed_fraction=None, max_failed_rows=None, row_timeout_seconds=None, init_timeout_seconds=None, num_workers=None, max_executor_processes_per_worker=None))]
#[pyo3(signature = (column, *, r#where=None, batch_size=None, max_concurrency=None, publication_interval_rows=None, resource_budget=None, supervisor_resources=None, planning_workers=None, gpu_class=None, gpu_count=None, retry_policy=None, error_policy=None, max_failed_fraction=None, max_failed_rows=None, row_timeout_seconds=None, init_timeout_seconds=None, num_workers=None, max_executor_processes_per_worker=None))]
#[allow(clippy::too_many_arguments)]
pub fn refresh_column_async<'py>(
self_: PyRef<'py, Self>,
@@ -2550,6 +2656,8 @@ impl Table {
max_concurrency: Option<Bound<'py, PyAny>>,
publication_interval_rows: Option<i64>,
resource_budget: Option<Bound<'py, PyDict>>,
supervisor_resources: Option<Bound<'py, PyDict>>,
planning_workers: Option<Bound<'py, PyDict>>,
gpu_class: Option<String>,
gpu_count: Option<i64>,
retry_policy: Option<String>,
@@ -2567,6 +2675,8 @@ impl Table {
max_concurrency: positive_u32_integer(max_concurrency.as_ref(), "max_concurrency")?,
publication_interval_rows,
resource_budget: parse_resource_budget(resource_budget.as_ref())?,
supervisor_resources: parse_supervisor_resources(supervisor_resources.as_ref())?,
planning_workers: parse_planning_workers(planning_workers.as_ref())?,
gpu_class,
gpu_count,
retry_policy,
+105 -1
View File
@@ -11560,6 +11560,59 @@ mod tests {
assert_eq!(job.id(), Some("j-budget"));
}
#[tokio::test]
async fn test_refresh_column_async_structured_resource_options() {
use crate::table::{
FunctionRefreshPlanningConfig, FunctionRefreshPodResources, RefreshColumnOptions,
};
let table = Table::new_with_handler("my_table", |request| {
assert_eq!(request.url().path(), "/v1/table/my_table/backfill_column");
assert_eq!(
request_body_json(&request),
serde_json::json!({
"column": "quality",
"supervisor_resources": {
"cpu_millis": 2_000,
"memory_bytes": 4_294_967_296_u64,
"ephemeral_storage_bytes": 10_737_418_240_u64,
},
"planning_workers": {
"resources": {
"cpu_millis": 1_000,
"memory_bytes": 2_147_483_648_u64,
"ephemeral_storage_bytes": 5_368_709_120_u64,
},
"max_workers": 4,
"rows_per_worker": 100_000,
"fragments_per_worker": 8,
},
})
);
http::Response::builder()
.status(202)
.body(r#"{"job_id":"j-structured"}"#)
.unwrap()
});
let supervisor = FunctionRefreshPodResources::new(2_000, 4_294_967_296, 10_737_418_240);
let planning = FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(1_000, 2_147_483_648, 5_368_709_120),
4,
100_000,
8,
);
let job = table
.refresh_column_async_with_options(
"quality",
RefreshColumnOptions::default()
.supervisor_resources(supervisor)
.planning_workers(planning),
)
.await
.unwrap();
assert_eq!(job.id(), Some("j-structured"));
}
#[tokio::test]
async fn test_refresh_column_async_retry_policy() {
use crate::function::{FunctionErrorPolicy, FunctionRetryPolicy};
@@ -11640,7 +11693,9 @@ mod tests {
#[tokio::test]
async fn test_refresh_column_rejects_invalid_filter_and_limit_options_before_request() {
use crate::table::RefreshColumnOptions;
use crate::table::{
FunctionRefreshPlanningConfig, FunctionRefreshPodResources, RefreshColumnOptions,
};
let table = Table::new_with_handler("my_table", |_| -> http::Response<String> {
panic!("invalid refresh options must not send a request")
@@ -11679,6 +11734,55 @@ mod tests {
RefreshColumnOptions::default().publication_interval_rows(0),
"publication_interval_rows",
),
(
RefreshColumnOptions::default()
.supervisor_resources(FunctionRefreshPodResources::new(0, 1, 1)),
"supervisor_resources.cpu_millis",
),
(
RefreshColumnOptions::default().planning_workers(
FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(1, 1, 0),
1,
1,
1,
),
),
"planning_workers.resources.ephemeral_storage_bytes",
),
(
RefreshColumnOptions::default().planning_workers(
FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(1, 1, 1),
0,
1,
1,
),
),
"planning_workers.max_workers",
),
(
RefreshColumnOptions::default().planning_workers(
FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(1, 1, 1),
1,
0,
1,
),
),
"planning_workers.rows_per_worker",
),
(
RefreshColumnOptions::default().planning_workers(
FunctionRefreshPlanningConfig::new(
FunctionRefreshPodResources::new(1, 1, 1),
1,
1,
0,
),
),
"planning_workers.fragments_per_worker",
),
] {
let error = table
.refresh_column_async_with_options("quality", options)
+4 -1
View File
@@ -120,7 +120,10 @@ pub use optimize::{
CompactionMode, CompactionOptions, CompactionOverrides, IndexRemapMode, OptimizeAction,
OptimizeStats,
};
pub use refresh::{FunctionWorkerResourceBudget, RefreshColumnOptions, RefreshColumnResult};
pub use refresh::{
FunctionRefreshPlanningConfig, FunctionRefreshPodResources, FunctionWorkerResourceBudget,
RefreshColumnOptions, RefreshColumnResult,
};
pub use rte_stats::{MemtableStats, RTEStats, SsTableStats, TableShardStats};
pub use schema_evolution::{
AddColumnsResult, AlterColumnsResult, DropColumnsResult, FieldMetadataUpdate,
+128
View File
@@ -104,6 +104,106 @@ impl FunctionWorkerResourceBudget {
}
}
/// CPU, memory, and ephemeral-storage limits for one Function refresh
/// control-plane pod.
///
/// This sizes the supervisor or planning worker pod for a registered
/// Function refresh. LanceDB Embedded rejects it.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(deny_unknown_fields)]
#[non_exhaustive]
pub struct FunctionRefreshPodResources {
/// CPU limit in millicores; 1000 is one CPU.
pub cpu_millis: u64,
/// Memory limit in bytes.
pub memory_bytes: u64,
/// Ephemeral storage limit in bytes.
pub ephemeral_storage_bytes: u64,
}
impl FunctionRefreshPodResources {
/// Create positive CPU, memory, and ephemeral-storage limits.
pub fn new(cpu_millis: u64, memory_bytes: u64, ephemeral_storage_bytes: u64) -> Self {
Self {
cpu_millis,
memory_bytes,
ephemeral_storage_bytes,
}
}
fn validate(&self, prefix: &str) -> Result<()> {
for (name, value) in [
("cpu_millis", self.cpu_millis),
("memory_bytes", self.memory_bytes),
("ephemeral_storage_bytes", self.ephemeral_storage_bytes),
] {
if value == 0 {
return Err(Error::InvalidInput {
message: format!("{prefix}.{name} must be greater than zero"),
});
}
}
Ok(())
}
}
/// Resources and sizing policy for the CPU-only planning fleet of one
/// registered Function refresh.
///
/// `resources` describes one planning worker pod. The remaining fields bound
/// how many workers may run and how much work each worker receives. LanceDB
/// Embedded rejects this option.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(deny_unknown_fields)]
#[non_exhaustive]
pub struct FunctionRefreshPlanningConfig {
/// CPU, memory, and ephemeral-storage limits for one planning worker.
pub resources: FunctionRefreshPodResources,
/// Maximum number of planning workers.
pub max_workers: usize,
/// Maximum rows assigned to one planning worker.
pub rows_per_worker: u64,
/// Maximum fragments assigned to one planning worker.
pub fragments_per_worker: u64,
}
impl FunctionRefreshPlanningConfig {
/// Create a planning configuration; every value must be positive.
pub fn new(
resources: FunctionRefreshPodResources,
max_workers: usize,
rows_per_worker: u64,
fragments_per_worker: u64,
) -> Self {
Self {
resources,
max_workers,
rows_per_worker,
fragments_per_worker,
}
}
fn validate(&self) -> Result<()> {
self.resources.validate("planning_workers.resources")?;
if self.max_workers == 0 {
return Err(Error::InvalidInput {
message: "planning_workers.max_workers must be greater than zero".into(),
});
}
if self.rows_per_worker == 0 {
return Err(Error::InvalidInput {
message: "planning_workers.rows_per_worker must be greater than zero".into(),
});
}
if self.fragments_per_worker == 0 {
return Err(Error::InvalidInput {
message: "planning_workers.fragments_per_worker must be greater than zero".into(),
});
}
Ok(())
}
}
/// Options for refreshing a computed column.
///
/// Serializes to the Enterprise backfill request body, one key per set
@@ -167,6 +267,14 @@ pub struct RefreshColumnOptions {
/// Function refresh. Cannot be combined with `gpu_count`.
#[serde(skip_serializing_if = "Option::is_none")]
pub resource_budget: Option<FunctionWorkerResourceBudget>,
/// CPU, memory, and ephemeral-storage limits for the Function refresh
/// supervisor pod. Enterprise only.
#[serde(skip_serializing_if = "Option::is_none")]
pub supervisor_resources: Option<FunctionRefreshPodResources>,
/// Resources and sizing limits for the Function refresh planning fleet.
/// Enterprise only.
#[serde(skip_serializing_if = "Option::is_none")]
pub planning_workers: Option<FunctionRefreshPlanningConfig>,
/// GPU class for Function workers in LanceDB Enterprise.
///
/// `None` keeps the Function's registered class, or the deployment default
@@ -264,6 +372,18 @@ impl RefreshColumnOptions {
self
}
/// Set the Function refresh supervisor pod resources.
pub fn supervisor_resources(mut self, resources: FunctionRefreshPodResources) -> Self {
self.supervisor_resources = Some(resources);
self
}
/// Set the Function refresh planning fleet resources and sizing.
pub fn planning_workers(mut self, planning: FunctionRefreshPlanningConfig) -> Self {
self.planning_workers = Some(planning);
self
}
/// Request Function workers from `gpu_class`.
pub fn gpu_class(mut self, gpu_class: impl Into<String>) -> Self {
self.gpu_class = Some(gpu_class.into());
@@ -297,6 +417,8 @@ impl RefreshColumnOptions {
self.publication_interval_rows.is_some(),
),
("resource_budget", self.resource_budget.is_some()),
("supervisor_resources", self.supervisor_resources.is_some()),
("planning_workers", self.planning_workers.is_some()),
]
.into_iter()
.find_map(|(name, set)| set.then_some(name))
@@ -367,6 +489,12 @@ impl RefreshColumnOptions {
if let Some(budget) = &self.resource_budget {
budget.validate()?;
}
if let Some(resources) = &self.supervisor_resources {
resources.validate("supervisor_resources")?;
}
if let Some(planning) = &self.planning_workers {
planning.validate()?;
}
if let Some(gpu_class) = &self.gpu_class {
validate_gpu_class_name(gpu_class)?;
}