Version: v0.66.1

Pump data

This example demonstrates how you can get information about all deployments in a specified project and start sending data to those deployments for scoring using the payload of their sample requests. This uses asynchronous scoring calls which makes it a good tool for testing a large number of deployments simultaneously.

Before you begin

Install aiohttp
You will need the values for the following constants in order to successfully carry out the task. Contact your administrator to obtain deployment specific values.

Constant	Value	Description
`MLOPS_API_URL`	Usually: `https://api.mlops.my.domain`	Defines the URL for the MLOps Gateway component.
`TOKEN_ENDPOINT_URL`	`https://mlops.keycloak.domain/auth/realms/[fill-in-realm-name]/protocol/openid-connect/token`	Defines the token endpoint URL of the Identity Provider. This uses Keycloak as the Identity Provider. Keycloak Realm should be provided.
`REFRESH_TOKEN`	`<your-refresh-token>`	Defines the user's refresh token.
`CLIENT_ID`	`<your-client-id>`	Sets the client id for authentication. This is the client you will be using to connect to MLOps.
`PROJECT_NAME`	`StandardFlow`	Defines a project name that the script will be using. This example uses the project created in Driverless AI example.
`SEND_INTERVAL`	`2.0`	Defines the time interval to send requests. This example sends sample requests every `2` seconds.
`MAX_CONSECUTIVE_ERRORS`	`5`	Stops sending requests to an endpoint after the defined number of errors. This example stops sending requests after `5` consecutive errors.
`CONNECTION_LIMIT`	`100`	Defines the maximum open TCP connections. This example defines a connection limit of `100`.

The following steps demonstrate how you can use MLOps Python client to send data to the deployments of a specified project for scoring.

Download the PumpData.py file.

Change the values of the following constants in your PumpData.py file as given in the preceding data table.

PumpData.py
### Constants
MLOPS_API_URL = <MLOPS_API_URL>
TOKEN_ENDPOINT_URL = <TOKEN_ENDPOINT_URL>
REFRESH_TOKEN = <REFRESH_TOKEN>
CLIENT_ID = <CLIENT_ID>
PROJECT_NAME = <PROJECT_NAME>
SEND_INTERVAL = <SEND_INTERVAL>
MAX_CONSECUTIVE_ERRORS = <MAX_CONSECUTIVE_ERRORS>
CONNECTION_LIMIT = <CONNECTION_LIMIT>

PumpData.py
### Constants
MLOPS_API_URL = "https://api.mlops.my.domain"
TOKEN_ENDPOINT_URL="https://mlops.keycloak.domain/auth/realms/[fill-in-realm-name]/protocol/openid-connect/token"
REFRESH_TOKEN="<your-refresh-token>"
CLIENT_ID="<your-mlops-client>"
PROJECT_NAME = "PumpDataExample"
SEND_INTERVAL = "2.0
MAX_CONSECUTIVE_ERRORS = 5
CONNECTION_LIMIT = 100

Run the PumpData.py file.
```
python3 PumpData.py
```
Navigate to MLOps and click the project name StandardFlow under Projects. Then click the name of the project to open the Deployment Overview side panel, and click view monitoring to see the monitoring details.
Note
For more information about deployments in MLOps, see Understand deployments in MLOps.
Under the Health panel you will be able to see the predictions made for the scoring data sent by the script.

Example walkthrough

This section provides a walkthrough of the PumpData.py file.

Include the following _get_project_id function to get the project id of the defined project name.

PumpData.py
def _get_project_id(projects: List[mlops.StorageProject]) -> str:
"""Gets the ID of the selected project."""
for p in projects:
    if p.display_name == PROJECT_NAME:
        return p.id
else:
    raise LookupError("Requested project not found")

Include the _get_sample_payload function to get a sample payload from the sample request deployment endpoint.

PumpData.py
async def _get_sample_payload(
    sample_request_url: str, deployment_id: str, http_session: aiohttp.ClientSession
) -> Dict[str, dict]:
    """Gets sample request payload from the sample request deployment endpoint."""
    async with http_session.get(sample_request_url) as r:
        sample_payload = await r.json()

    return {deployment_id: sample_payload}

Include the get_deployment_statuses function to get the deployment statuses of the project. This first gets all the projects of the user and then gets the deployment statuses of the project the user has specified in the script.

PumpData.py
def get_deployment_statuses(client: mlops.Client) -> List[mlops.DeployDeploymentStatus]:
    """Gets deployment metada."""
    # Getting all user's projects.
    projects: mlops.StorageListProjectsResponse = client.storage.project.list_projects(
        mlops.StorageListProjectsRequest()
    )
    prj_id = _get_project_id(projects.project)

    # Getting all deployment statuses for the selected project.
    statuses: mlops.DeployListDeploymentStatusesResponse = (
        client.deployer.deployment_status.list_deployment_statuses(
            mlops.DeployListDeploymentStatusesRequest(project_id=prj_id)
        )
    )

    return statuses.deployment_status

Include the get_sample_payloads function to get sample payloads from a collection of deployments metadata.

PumpData.py
async def get_sample_payloads(
    deployment_statuses: List[mlops.DeployDeploymentStatus],
    http_session: aiohttp.ClientSession,
) -> dict:
    """Gets sample payloads from a collection of deployments metadata."""
    requests_futures = []
    for ds in deployment_statuses:
        sample_payload_awaitable = _get_sample_payload(
            ds.scorer.sample_request.url, ds.deployment_id, http_session
        )
        sample_payload_future = asyncio.ensure_future(sample_payload_awaitable)
        requests_futures.append(sample_payload_future)
    sample_payloads_nested = await asyncio.gather(*requests_futures)

    sample_payloads = {}
    for payload in sample_payloads_nested:
        sample_payloads.update(payload)

    return sample_payloads

Include the send_sample_request function to continuously sends sample requests to the specified scoring URL.

PumpData.py
async def send_sample_request(
    scoring_url: str, payload: dict, http_session: aiohttp.ClientSession
) -> None:
    """Continously sends sample requests to the specified scoring URL."""
    error_count = 0
    while error_count < MAX_CONSECUTIVE_ERRORS:
        async with http_session.post(scoring_url, json=payload) as r:
            if r.status != 200:
                error_count += 1
                print(
                    f"Failed sending data to: {scoring_url} with status: {r.status}, response: {await r.text()}"
                )
        await asyncio.sleep(SEND_INTERVAL)
    print(f"Stopped sending requests to: {scoring_url}")

In the main function,

Set up the TCP connection limit.

PumpData.py
conn = aiohttp.TCPConnector(limit=CONNECTION_LIMIT)

Set up the token provider using an existing refresh token.
Set up the MLOps client.
Get the deployment metadata by calling the get_deployment_statuses function.
PumpData.py
```
deployment_statuses = get_deployment_statuses(mlops_client)
```

Include the client session.

PumpData.py
async with aiohttp.ClientSession(connector=conn) as sessions:
    sample_payloads = await get_sample_payloads(deployment_statuses, sessions)
    sample_requests_futures = []
    for ds in deployment_statuses:
        send_request_awaitable = send_sample_request(
            ds.scorer.score.url, sample_payloads[ds.deployment_id], sessions
        )
        send_request_future = asyncio.ensure_future(send_request_awaitable)
        sample_requests_futures.append(send_request_future)
    await asyncio.gather(*sample_requests_futures)

Note

Although having a single client session is uncommon, this context manager is actually a connection pool and should be treated as such. Having multiple client sessions here can cause race condition issues. For more information, see Client session in aiohttp.

Finally, include the below code to run the main function.

PumpData.py
if __name__ == "__main__":
loop = asyncio.get_event_loop()
loop.run_until_complete(main())

Feedback

Submit and view feedback for this page
Send feedback about H2O MLOps to cloud-feedback@h2o.ai

Pump data

Example walkthrough​

Example walkthrough