lightbeam 0.1.6

Sends JSONL data into an Ed-Fi API

lightbeam transmits payloads from JSONL files into an Ed-Fi API.

Table of Contents

• Requirements
• Installation
• Setup
• Usage
• Features
• Design
• Performance
• Limitations
• Changelog
• Contributing
• License

Requirements

Python 3, pip, and connectivity to an Ed-Fi API.

Installation

pip install lightbeam

Setup

Running the tool requires

1. a folder of JSONL files, one for each Ed-Fi Resource and Descriptor to populate
2. a YAML configuration file An example YAML configuration is below, followed by documentation of each option.

state_dir: ~/.lightbeam/
data_dir: ./
namespace: ed-fi
edfi_api:
  base_url: https://api.schooldistrict.org/v5.3/api
  oauth_url: https://api.schooldistrict.org/v5.3/api/oauth/token 
  dependencies_url: https://api.schooldistrict.org/v5.3/api/metadata/data/v3/2024/dependencies
  open_api_metadata_url: https://api.schooldistrict.org/v5.3/api/metadata/
  descriptors_swagger_url: https://api.schooldistrict.org/v5.3/api/metadata/data/v3/2024/descriptors/swagger.json
  resources_swagger_url: https://api.schooldistrict.org/v5.3/api/metadata/data/v3/2024/resources/swagger.json
  version: 3
  mode: year_specific
  year: 2021
  client_id: yourID
  client_secret: yourSecret
connection:
  pool_size: 8
  timeout: 60
  num_retries: 10
  backoff_factor: 1.5
  retry_statuses: [429, 500, 502, 503, 504]
  verify_ssl: True
count:
  separator: ,
fetch:
  page_size: 100
force_delete: True
log_level: INFO
show_stacktrace: True

• (optional) state_dir is where state is stored. The default is ~/.lightbeam/ on *nix systems, C:/Users/USER/.lightbeam/ on Windows systems.
• (optional) Specify the data_dir which contains JSONL files to send to Ed-Fi. The default is ./. The tool will look for files like {Resource}.jsonl or {Descriptor}.jsonl in this location, as well as directory-based files like {Resource}/*.jsonl or {Descriptor}/*.jsonl. Files with .ndjson or simply .json extensions will also be processed. (More info at the ndjson standard page.)
• (optional) Specify the namespace to use when accessing the Ed-Fi API. The default is ed-fi but others include tpdm or custom values. To send data to multiple namespaces, you must use a YAML configuration file and lightbeam send for each.
• Specify the details of the edfi_api to which to connect including
  • (optional) The base_url which serves a JSON object specifying the paths to data endpoints, Swagger, and dependencies. The default is https://localhost/api (the address of an Ed-Fi API running locally in Docker), but the location varies depending on how Ed-Fi is deployed.
  • Most Ed-Fi APIs publish other endpoints they provide in a JSON document at the base URL - lighbteam attempts to discover these. If, however, your API does not publish these URLs at the base, or you want to manually override any of them for any reason, you can specify the following:
    • (optional) oauth_url (usually [base_url]/oauth/token)
    • (optional) dependencies_url (usually [base_url]/metadata/data/v3/dependencies)
    • (optional) descriptors_swagger_url (usually [base_url]/metadata/data/v3/descriptors/swagger.json)
    • (optional) resources_swagger_url (usually [base_url]/metadata/data/v3/resources/swagger.json)
    • (optional) open_api_metadata_url (usually [base_url]/metadata/) - this is ignored if both descriptors_swagger_url and resources_swagger_url are specified
  • The version as one of 3 or 2 (2 is currently unsupported).
  • (optional) The mode as one of shared_instance, sandbox, district_specific, year_specific, or instance_year_specific.
  • (required if mode is year_specific or instance_year_specific) The year used to build the resource URL. The default is the current year.
  • (required if mode is instance_year_specific) The instance_code used to build the resource URL. The default is none.
  • (required) Specify the client_id to use when connecting to the Ed-Fi API.
  • (required) Specify the client_secret to use when connecting to the Ed-Fi API.
• Specify the connection parameters to use when making requests to the API including
  • (optional) The pool_size. The default is 8. The optimal setting depends on the Ed-Fi API's capabilities.
  • (optional) The timeout (in seconds) to wait for each connection attempt. The default is 60 seconds.
  • (optional) The num_retries to do in case of request failures. The default is 10.
  • (optional) The backoff_factor to use for the exponential backoff. The default is 1.5.
  • (optional) The retry_statuses, that is, the HTTPS response codes to consider as failures to retry. The default is [429, 500, 501, 503, 504].
  • (optional) Whether to verify_ssl. The default is True. Set to False when working with localhost APIs or to live dangerously.
• (optional) for lightbeam count, optionally change the separator between Records and Endpoint. The default is a "tab" character.
• (optional) for lightbeam fetch, optionally specify the number of records (page_size) to GET at a time. The default is 100, but if you're trying to extract lots of data from an API increase this to the largest allowed (which depends on the API, but is often 500 or even 5000).
• (optional) Skip the interactive confirmation prompt (for programmatic use) when using the delete command. The default is False (prompt).
• (optional) Specify a log_level for output. Possible values are
  • ERROR: only output errors like missing required sources, invalid references, invalid YAML configuration, etc.
  • WARNING: output errors and warnings like when the run log is getting long
  • INFO: all errors and warnings plus basic information about what earthmover is doing: start and stop, how many rows were removed by a distinct_rows or filter_rows operation, etc. (This is the default log_level.)
  • DEBUG: all output above, plus verbose details about each transformation step, timing, memory usage, and more. (This log_level is recommended for debugging transformations.)
• (optional) Specify whether to show a stacktrace for runtime errors. The default is False.

Usage

lightbeam recognizes several commands:

count

lightbeam count -c path/to/config.yaml

Prints to the console (or to your --results-file, if specified) a record count for each endpoint in your Ed-Fi API.

• By default, resources and descriptors (all endpoints) are counted. You can change this by using selectors, such as -e *Descriptors.
• Endpoint counts printed to the console (if you don't specify a --results-file) include only endpoints with more than zero records. Endpoint counts saved in a --results-file include all available endpoints, even those with zero records.
• Whether printed to the console or a --results-file, output will include columns Records and Endpoint separated by a separator specified as count.separator in your YAML configuration (default is a "tab" character).

fetch

lightbeam fetch -c path/to/config.yaml

Fetches the payloads of selected endpoints from your Ed-Fi API and saves them, each on their own line, to JSONL files in your data_dir.

Optionally specify --query '{"studentUniqueId": 12345}' or -q '{"key": "value"}' to add query parameters to every GET request. This can be useful if you want to fetch data for just a specific record (and related data). For example:

lightbeam fetch -s student* -e *Descriptors -q '{"studentUniqueId":12345}' -d id,_etag,_lastModifiedDate

Optionally specify --keep-keys id or -k id to keep only specific keys from every payload. This can be useful to reduce the amount of data stored if you only need certain fields. It is used internally by truncate to only fetch the ids or payloads to then delete by id.

Optionally specify --drop-keys id,_etag,_lastModified or -d id to remove specific keys from every payload. This can be useful if you want to fetch data from one Ed-Fi API and then turn around and send it to another.

Like selectors, keep-keys and drop-keys are comma-separated lists of values, each of which may begin or end with an asterisk (*) for wildcard matching. Example: -d _* would remove properties beginning with an underscore (_) character from any fetched payloads.

validate

lightbeam validate -c path/to/config.yaml

You may validate your JSONL before transmitting it. Configuration for validate goes in its own section of lightbeam.yaml:

validate:
  methods:
    - schema # checks that payloads conform to the Swagger definitions from the API
    - descriptors # checks that descriptor values are either locally-defined or exist in the remote API
    - uniqueness # checks that local payloads are unique by the required property values
    - references # checks that references resolve, either locally or in the remote API
  # or
  # methods: "*"

Default validate.methods are ["schema", "descriptors", "uniqueness"] (not references; see below). In addition to the above methods, lighteam validate will also (first) check that each payload is valid JSON.

The references method can be slow, as a separate GET request may be made to your API for each reference. (Therefore the validation method is disabled by default.) lightbeam tries to improve efficiency by:

• batching requests and sending several concurrently (based on connection.pool_size of lightbeam.yaml)
• caching responses and first checking the cache before making another (potentially identical) request

Even with these optimizations, checking references can easily take minutes for even relatively small amounts of data. Therefore lightbeam.yaml also accepts a further configuration option:

validate:
  references:
    max_failures: 10 # stop testing after X failed payloads ("fail fast")

This is optional; if absent, references in every payload are checked, no matter how many fail.

Note: Reference validation efficiency may be improved by first lightbeam fetching certain resources to have a local copy. lightbeam validate checks local JSONL files to resolve references before trying the remote API, and fetch retrieves many records per GET, so total runtime can be faster in this scenario. The downsides include

• more data movement
• fetched data becoming stale over time
• needing to track which data is your own vs. was fetched (all the data must coexist in the config.data_dir to be discoverable by lightbeam validate)

send

lightbeam send -c path/to/config.yaml

Sends your JSONL payloads to your Ed-Fi API.

validate+send

lightbeam validate+send -c path/to/config.yaml

This is a shorthand for sequentially running validate and then send. It can be useful to catching errors in automated pipelines earlier in the validate step before you actually send problematic data to your Ed-Fi API.

delete

lightbeam delete -c path/to/config.yaml

Delete payloads by

1. determing the natural key (set of required fields) for each endpoint
2. iterating through your JSONL payloads and looking up each one via a GET request to the API filtering for the natural key values
3. if exactly one result is returned, DELETEing it by id

Payload hashes are also deleted from saved state. Endpoints are processed in reverse-dependency order to prevent delete failures due to data dependencies.

Note that the default profile for most Ed-Fi API credentials prevents deletion of certain core resources (student, school, etc.), even if your credentials were used to create the records. If you get API errors trying to delete records, you may need "no further auth" API credentials.

Running the delete command will prompt you to type "yes" to confirm. This confirmation prompt can be disabled (for programmatic use) by specifying force_delete: True in your YAML.

truncate

lightbeam truncate -c path/to/config.yaml

Truncates (empties) your Ed-Fi API for selected endpoints, in dependency-order. USE WITH CAUTION! truncate works by fetching the id of every record for a given endpoint and then deleting all records by ID.

Truncateing a resource will also clear out the saved state for it.

Note that the default profile for most Ed-Fi API credentials prevents deletion of certain core resources (student, school, etc.), even if your credentials were used to create the records. If you get API errors trying to delete records, you may need "no further auth" API credentials.

Running the truncate command will prompt you to type "yes" to confirm. This confirmation prompt can be disabled (for programmatic use) by specifying force_delete: True in your YAML.

truncate is a convenience command which should be used sparingly, as it can generate large numbers of deletes records and cause performance issues when pulling from deletes endpoints. If you want to wipe an entire Ed-Fi ODS, a better approach may be to drop and recreate the database (and re-send Descriptors and other default resources as needed).

Other options

See a help message with

lightbeam -h
lightbeam --help

See the tool version with

lightbeam -v
lightbeam --version

Features

This tool includes several special features:

Selectors

Send only a subset of resources or descriptors in your data_dir using -s or --selector:

lightbeam send -c path/to/config.yaml -s schools,students,studentSchoolAssociations

or, similarly, exclude some resources or descriptors using -e or --exclude:

lightbeam send -c path/to/config.yaml -e *Descriptors

Selection and exclusion may be a single or comma-separated list of strings or a wildcards (beginning or ending with *). For example:

lightbeam send -c path/to/config.yaml -s student*,parent* -e *Associations,*Descriptors

would process resources like studentSchoolAttendanceEvents and parents, but not studentSchoolAssociations, studentParentAssociations, or any Descriptors.

Environment variable references

In your YAML configuration, you may reference environment variables with ${ENV_VAR}. This can be useful for passing sensitive data like credentials to lightbeam, such as

...
edfi_api:
  client_id: ${EDFI_API_CLIENT_ID}
  client_secret: ${EDFI_API_CLIENT_SECRET}
...

Command-line parameters

Similarly, you can specify parameters via the command line with

lightbeam send -c path/to/config.yaml -p '{"CLIENT_ID":"populated", "CLIENT_SECRET":"populatedSecret"}'
lightbeam send -c path/to/config.yaml --params '{"CLIENT_ID":"populated", "CLIENT_SECRET":"populatedSecret"}'

Command-line parameters override any environment variables of the same name.

State

This tool maintains state about payloads previously dispatched to the Ed-Fi API to avoid repeatedly resending the same payloads. This is done by maintaining a pickled Python dictionary of payload hashes for each Ed-Fi resource and descriptor, together with a timestamp and HTTP status code of the last response. The files are located in the config file's state_dir and have names like {resource}.dat or {descriptor}.dat.

By default, only new, never-before-seen payloads are sent or deleted.

You may choose to resend payloads last sent before timestamp using the -t or --older-than command-line flag:

lightbeam send -c path/to/lightbeam.yaml -t 2020-12-25T00:00:00
lightbeam send -c path/to/lightbeam.yaml --older-than 2020-12-25T00:00:00

Or you may choose to resend payloads last sent after timestamp using the -n or --newer-than command-line flag:

lightbeam send -c path/to/lightbeam.yaml -n 2020-12-25T00:00:00
lightbeam send -c path/to/lightbeam.yaml --newer-than 2020-12-25T00:00:00

Or you may choose to resend payloads that returned a certain HTTP status code(s) on the last send using the -r or --retry-status-codes command-line flag:

lightbeam send -c path/to/lightbeam.yaml -r 200,201
lightbeam send -c path/to/lightbeam.yaml --retry-status-codes 200,201

These three options may be composed; lightbeam will resend payloads that match any conditions (logical OR).

Finally, you can ignore prior state and resend all payloads using the -f or --force flag:

lightbeam send -c path/to/lightbeam.yaml -f
lightbeam send -c path/to/lightbeam.yaml --force

Cache

To reduce runtime, lightbeam caches the resource and descriptor Swagger docs it fetches from your Ed-Fi API as well as the descriptor values for up to a month. This way, the data does not have to be re-loaded from your API on every run. The cached files are stored in the cache directory within your state_dir. You may run lightbeam with the -w or --wipe flag to clear this cached data and force re-fetching the API metadata:

lightbeam send -c path/to/config.yaml -w
lightbeam send -c path/to/config.yaml --wipe

Structured output of run results

To produce a JSON file with metadata about the run, invoke lightbeam with

lightbeam send -c path/to/config.yaml --results-file ./results.json

A sample results file could be:

{
    "started_at": "2023-06-08T17:18:25.053207",
    "working_dir": "/home/someuser/",
    "config_file": "/home/someuser/lightbeam.yml",
    "data_dir": "/home/someuser/data/",
    "api_url": "https://some-ed-fi-api.edu/api",
    "namespace": "ed-fi",
    "resources": {
        "studentSchoolAssociations": {
            "records_processed": 50,
            "records_skipped": 0,
            "records_failed": 22,
            "failures": [
                {
                    "status_code": 400,
                    "message": "The request is invalid. \"request.schoolReference.schoolId\": [ \"JSON integer 1234567899999 is too large or small for an Int32. Path 'schoolReference.schoolId', line 1, position 328.\" ]",
                    "file": "/home/someuser/data/studentSchoolAssociations.jsonl",
                    "line_numbers": "6,4,5,7,8",
                    "count": 5
                },
                {
                    "status_code": 400,
                    "message": "Validation of 'StudentSchoolAssociation' failed. Student reference could not be resolved.",
                    "file": "/home/someuser/data/studentSchoolAssociations.jsonl",
                    "line_numbers": "1,3,2",
                    "count": 3

                },
                {
                    "status_code": 409,
                    "message": "The value supplied for the related 'studentschoolassociation' resource does not exist."
                    "file": "/home/someuser/data/studentSchoolAssociations.jsonl",
                    "line_numbers": "9,10,12,14,16,13,11,15,17,18,19,21,22,20",
                    "count": 14
                }
            ],
            "successes": [
                {
                    "status_code": 201,
                    "count": 28
                }
            ],
            "records_processed": 50,
            "records_skipped": 0,
            "records_failed": 22
        }
    },
    "completed_at": "2023-06-08T17:18:26.724699",
    "runtime_sec": 1.671492,
    "total_records_processed": 50,
    "total_records_skipped": 0,
    "total_records_failed": 22
}

Design

Some details of the design of this tool are discussed below.

Resource-dependency ordering

JSONL files are sent to the Ed-Fi API in resource-dependency order, which avoids "missing reference" API errors when populating multiple endpoints.

Asynchronous requests

lightbeam achieves exceptional performance by making asynchronous requests to the Ed-Fi API - up to connection.pool_size (in your YAML configuration) at a time.

Performance & Limitations

Tool performance depends on primarily on the performance of the Ed-Fi API, which in turn depends on the compute resources which back it. Typically the bottleneck is write performance to the database backend (SQL server or Postgres). If you use lightbeam to ingest a large amount of data into an Ed-Fi API (not a recommended use-case), consider temporarily scaling up your database backend.

For reference, we have achieved throughput rates in excess of 100 requests/second against an Ed-Fi ODS & API running in Docker on a laptop.

Changelog

See CHANGELOG.

Contributing

Bugfixes and new features (such as additional transformation operations) are gratefully accepted via pull requests here on GitHub.

Contributions

• Cover image created with DALL • E mini

License

See License.