Copyright (C) 2018-2022 The Open Library Foundation
This software is distributed under the terms of the Apache License, Version 2.0. See the file "LICENSE" for more information.
- Introduction
- Compiling
- Docker
- Installing the module
- Deploying the module
- Maximum upload file size and java heap memory setups
- Scalability
- File splitting configuration
- Interaction with AWS S3/Minio
- Queue prioritization algorithm
- Interaction with Kafka
- Other system properties
- Issue tracker
- Additional information
- Script to upload a batch of MARC records
mod-data-import is responsible for uploading files (see documentation for file uploading), initial handling and sending records for further processing (see documentation for file processing).
mvn install
See that it says "BUILD SUCCESS" near the end.
Build the docker container with:
docker build -t mod-data-import .
Test that it runs with:
docker run -t -i -p 8081:8081 mod-data-import
Follow the guide of Deploying Modules sections of the Okapi Guide and Reference, which describe the process in detail.
First of all you need a running Okapi instance. (Note that specifying an explicit 'okapiurl' might be needed.)
cd .../okapi
java -jar okapi-core/target/okapi-core-fat.jar dev
We need to declare the module to Okapi:
curl -w '\n' -X POST -D - \
-H "Content-type: application/json" \
-d @target/ModuleDescriptor.json \
http://localhost:9130/_/proxy/modules
That ModuleDescriptor tells Okapi what the module is called, what services it provides, and how to deploy it.
Next we need to deploy the module. There is a deployment descriptor in
target/DeploymentDescriptor.json. It tells Okapi to start the module on 'localhost'.
Deploy it via Okapi discovery:
curl -w '\n' -D - -s \
-X POST \
-H "Content-type: application/json" \
-d @target/DeploymentDescriptor.json \
http://localhost:9130/_/discovery/modules
Then we need to enable the module for the tenant:
curl -w '\n' -X POST -D - \
-H "Content-type: application/json" \
-d @target/TenantModuleDescriptor.json \
http://localhost:9130/_/proxy/tenants/<tenant_name>/modules
Current implementation supports only storing of the file in a LOCAL_STORAGE (file system of the module). It has a couple of implications:
- the request for processing the file can be processed only by the same instance of the module, which prevents mod-data-import from scaling
- file size that can be uploaded is limited to the java heap memory allocated to the module. It is necessary to have the size of the java heap equal to the expected max file size plus 10 percent.
| File Size | Java Heap size |
|---|---|
| 256mb | 270+ mb |
| 512mb | 560+ mb |
| 1GB | 1.1+ GB |
To initialise processing of a file user should choose a Job Profile - that information is crucial as it basically contains the instructions on what to do with the uploaded file. However, this process happens after file is uploaded and comes to mod-data-import as a separate request. External storage is required to make mod-data-import scalable. Implementation of the module has the possibility to read the configuration settings from mod-configuration. To allow multiple instance deployment, for every instance the same persistent volume must be mounted to the mount point defined by the value of data.import.storage.path property.
- data.import.storage.type - type of data storage used for uploaded files. Default value is LOCAL_STORAGE. Other implementations for storage should be added.
- data.import.storage.path - path where uploaded file will be stored
The file-splitting process may be configured with the following environment variables:
| Name | Type | Required | Default | Description |
|---|---|---|---|---|
SPLIT_FILES_ENABLED |
true or false |
yes, if enabling feature | false |
Whether files should be split into chunks and processed separately |
RECORDS_PER_SPLIT_FILE |
integer > 0 | no | 1000 |
The maximum number of records to include in a single file |
ASYNC_PROCESSOR_POLL_INTERVAL_MS |
integer (msec) ≥ 0 | no | 5000 |
The number of milliseconds between times when the module checks the queue for waiting jobs |
ASYNC_PROCESSOR_MAX_WORKERS_COUNT |
integer ≥ 1 | no | 1 |
The maximum number of concurrent jobs to process at once, in this instance |
For the polling interval, a lower number results in decreased latency between when a job is added to the queue and when it is processed. However, this also results in more frequent database queries, which may impact performance. Note that the number set here is the "worst case" — average waiting would be half of it — and that a few seconds delay on a large import is hardly noticeable.
The worker count is useful for production/multi-tenant environments, where you might want to provide more capacity without additional instances. However, note that this may cause some odd behavior when only one user is running a job, as multiple parts may appear to complete together.
Note
For full information about this feature, please view the release notes
This module uses S3-compatible storage as part of the file upload process. The following environment variables must be set with values for your S3-compatible storage (AWS S3, Minio Server):
| Name | Type | Required | Default | Description |
|---|---|---|---|---|
AWS_URL |
URL as string | yes | http://127.0.0.1:9000/ |
URL of S3-compatible storage |
AWS_REGION |
string | yes | none | S3 region |
AWS_BUCKET |
string | yes | none | Bucket to store and retrieve data |
AWS_ACCESS_KEY_ID |
string | yes | none | S3 access key |
AWS_SECRET_ACCESS_KEY |
string | yes | none | S3 secret key |
AWS_SDK |
true or false |
no, if using MinIO | false |
If AWS S3 is being used (true if so, false other platforms such as MinIO) |
S3_FORCEPATHSTYLE |
true or false |
no | false |
If path-style requests should be used instead of virtual-hosted style |
Path-style vs virtual-hosted style requests are described on the AWS S3 documentation.
Warning
It is possible for files to be partially uploaded but abandoned in the UI. This module makes no effort to detect these cases and proactively delete them.
Instead, use the retention policies built into AWS S3 and MinIO, as described here.
This covers the following environment variables:
Note
None of these are required; if not set, the following default values will be used.
| Name | Type (unit) | Default | Reasoning |
|---|---|---|---|
SCORE_JOB_SMALLEST |
integer | 40 |
|
SCORE_JOB_LARGEST |
integer | -40 |
Larger jobs should be deprioritized |
SCORE_JOB_REFERENCE |
integer (records) | 100000 |
|
SCORE_AGE_NEWEST |
integer | 0 |
New jobs begin with no boost |
SCORE_AGE_OLDEST |
integer | 50 |
As jobs age, their score increases rapidly, so this does not have to be too high. We want small jobs to "cut" in line effectively. |
SCORE_AGE_EXTREME_THRESHOLD_MINUTES |
integer (minutes) | 480 |
8 hours |
SCORE_AGE_EXTREME_VALUE |
integer | 10000 |
Jump to the top of the queue if waiting more than 8 hours |
SCORE_TENANT_USAGE_MIN |
integer | 100 |
If the tenant has no jobs running, then it should be prioritized |
SCORE_TENANT_USAGE_MAX |
integer | -200 |
If the tenant is using all available workers, it should be significantly deprioritized. If no other tenants are competing, this will not matter (since all jobs would be offset by this) |
SCORE_PART_NUMBER_FIRST |
integer | 1 |
Very small; we only want to order parts amongst others within a job (which would likely have the same score otherwise) |
SCORE_PART_NUMBER_LAST |
integer | 0 |
|
SCORE_PART_NUMBER_LAST_REFERENCE |
integer | 100 |
Does not really matter due to the small range |
For information on what these mean, how to configure them, how scores are calculated, and even a playground to try experiment with different values, please see this wiki page.
Important
To disable an individual metric (or the prioritization altogether), set the value(s) to 0.
Note
We recommend the suggested values above, however, there is a lot of room for customization and extension as needed. Please see the doc for more information.
All modules involved in data import (mod-data-import, mod-source-record-manager, mod-source-record-storage, mod-inventory, mod-invoice) are communicating via Kafka directly. Therefore, to enable data import Kafka should be set up properly and all the necessary parameters should be set for the modules.
Properties that are required for mod-data-import to interact with Kafka:
KAFKA_HOSTKAFKA_PORTOKAPI_URLENV(unique env ID)
There are another important properties - number of partitions for topics DI_INITIALIZATION_STARTED and DI_RAW_RECORDS_CHUNK_READ
which are created during tenant initialization, the values of which can be customized with
DI_INITIALIZATION_STARTED_PARTITIONS and DI_RAW_RECORDS_CHUNK_READ_PARTITIONS env variables respectively.
Default value - 1.
Initial handling of the uploaded file means chunking it and sending records for processing in other modules. The chunk size can be adjusted for different files, otherwise default values will be used:
- "file.processing.marc.raw.buffer.chunk.size": 50 - applicable to MARC files in binary format
- "file.processing.marc.json.buffer.chunk.size": 50 - applicable to json files with MARC data in json format
- "file.processing.marc.xml.buffer.chunk.size": 10 - applicable to xml files with MARC data in xml format
- "file.processing.edifact.buffer.chunk.size": 10 - applicable to EDIFACT files
See project MODDATAIMP at the FOLIO issue tracker.
The raml-module-builder framework.
Other modules.
See project MODDATAIMP at the FOLIO issue tracker.
Other FOLIO Developer documentation is at dev.folio.org
The scripts directory contains a shell-script, load-marc-data-into-folio.sh, and a file with a sample of 100 MARC records, sample100.marc. This script can be used to upload any batch of MARC files automatically, using the same sequence of WSAPI operations as the Secret Button. First, login to a FOLIO backend service using the Okapi command-line utility or any other means that leaves definitions of the Okapi URL, tenant and token in the .okapi file in the home directory. Then run the script, naming the MARC file as its own argument:
scripts$ echo OKAPI_URL=https://folio-snapshot-stable-okapi.dev.folio.org > ~/.okapi
scripts$ echo OKAPI_TENANT=diku >> ~/.okapi
scripts$ okapi login
username: diku_admin
password: ************
Login successful. Token saved to /Users/mike/.okapi
scripts$ ./load-marc-data-into-folio.sh sample100.marc
=== Stage 1 ===
=== Stage 2 ===
=== Stage 3 ===
HTTP/2 204
date: Thu, 27 Aug 2020 11:55:28 GMT
x-okapi-trace: POST mod-authtoken-2.6.0-SNAPSHOT.73 http://10.36.1.38:9178/data-import/uploadDefinitions/123a8d01-e389-4893-a53e-cc2de846471d/processFiles.. : 202 7078us
x-okapi-trace: POST mod-data-import-1.11.0-SNAPSHOT.140 http://10.36.1.38:9175/data-import/uploadDefinitions/123a8d01-e389-4893-a53e-cc2de846471d/processFiles.. : 204 6354us
scripts$