Async crawler and datalake service for data.gouv.fr
Project description
udata-hydra 🦀
udata-hydra is an async metadata crawler for data.gouv.fr.
URLs are crawled via aiohttp, catalog and crawled metadata are stored in a PostgreSQL database.
CLI
Create database structure
Install udata-hydra dependencies and cli.
poetry install
poetry run udata-hydra migrate
Load (UPSERT) latest catalog version from data.gouv.fr
udata-hydra load-catalog
Crawler
udata-hydra-crawl
It will crawl (forever) the catalog according to config set in config.py.
BATCH_SIZE URLs are queued at each loop run.
The crawler will start with URLs never checked and then proceed with URLs crawled before SINCE interval. It will then wait until something changes (catalog or time).
There's a by-domain backoff mecanism. The crawler will wait when, for a given domain in a given batch, BACKOFF_NB_REQ is exceeded in a period of BACKOFF_PERIOD seconds. It will retry until the backoff is lifted.
If an URL matches one of the EXCLUDED_PATTERNS, it will never be checked.
Worker
A job queuing system is used to process long-running tasks. Launch the worker with the following command:
poetry run rq worker -c udata_hydra.worker
API
Run
poetry install
poetry run adev runserver udata_hydra/app.py
Get latest check
Works with ?url={url} and ?resource_id={resource_id}.
$ curl -s "http://localhost:8000/api/checks/latest/?url=http://opendata-sig.saintdenis.re/datasets/661e19974bcc48849bbff7c9637c5c28_1.csv" | json_pp
{
"status" : 200,
"catalog_id" : 64148,
"deleted" : false,
"error" : null,
"created_at" : "2021-02-06T12:19:08.203055",
"response_time" : 0.830198049545288,
"url" : "http://opendata-sig.saintdenis.re/datasets/661e19974bcc48849bbff7c9637c5c28_1.csv",
"domain" : "opendata-sig.saintdenis.re",
"timeout" : false,
"id" : 114750,
"dataset_id" : "5c34944606e3e73d4a551889",
"resource_id" : "b3678c59-5b35-43ad-9379-fce29e5b56fe",
"headers" : {
"content-disposition" : "attachment; filename=\"xn--Dlimitation_des_cantons-bcc.csv\"",
"server" : "openresty",
"x-amz-meta-cachetime" : "191",
"last-modified" : "Wed, 29 Apr 2020 02:19:04 GMT",
"content-encoding" : "gzip",
"content-type" : "text/csv",
"cache-control" : "must-revalidate",
"etag" : "\"20415964703d9ccc4815d7126aa3a6d8\"",
"content-length" : "207",
"date" : "Sat, 06 Feb 2021 12:19:08 GMT",
"x-amz-meta-contentlastmodified" : "2018-11-19T09:38:28.490Z",
"connection" : "keep-alive",
"vary" : "Accept-Encoding"
}
}
Get all checks for an URL or resource
Works with ?url={url} and ?resource_id={resource_id}.
$ curl -s "http://localhost:8000/api/checks/all/?url=http://www.drees.sante.gouv.fr/IMG/xls/er864.xls" | json_pp
[
{
"domain" : "www.drees.sante.gouv.fr",
"dataset_id" : "53d6eadba3a72954d9dd62f5",
"timeout" : false,
"deleted" : false,
"response_time" : null,
"error" : "Cannot connect to host www.drees.sante.gouv.fr:443 ssl:True [SSLCertVerificationError: (1, \"[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: Hostname mismatch, certificate is not valid for 'www.drees.sante.gouv.fr'. (_ssl.c:1122)\")]",
"catalog_id" : 232112,
"url" : "http://www.drees.sante.gouv.fr/IMG/xls/er864.xls",
"headers" : {},
"id" : 165107,
"created_at" : "2021-02-06T14:32:47.675854",
"resource_id" : "93dfd449-9d26-4bb0-a6a9-ee49b1b8a4d7",
"status" : null
},
{
"timeout" : false,
"deleted" : false,
"response_time" : null,
"error" : "Cannot connect to host www.drees.sante.gouv.fr:443 ssl:True [SSLCertVerificationError: (1, \"[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: Hostname mismatch, certificate is not valid for 'www.drees.sante.gouv.fr'. (_ssl.c:1122)\")]",
"domain" : "www.drees.sante.gouv.fr",
"dataset_id" : "53d6eadba3a72954d9dd62f5",
"created_at" : "2020-12-24T17:06:58.158125",
"resource_id" : "93dfd449-9d26-4bb0-a6a9-ee49b1b8a4d7",
"status" : null,
"catalog_id" : 232112,
"url" : "http://www.drees.sante.gouv.fr/IMG/xls/er864.xls",
"headers" : {},
"id" : 65092
}
]
Get crawling status
$ curl -s "http://localhost:8000/api/status/crawler/" | json_pp
{
"fresh_checks_percentage" : 0.4,
"pending_checks" : 142153,
"total" : 142687,
"fresh_checks" : 534,
"checks_percentage" : 0.4
}
Get worker status
$ curl -s "http://localhost:8000/api/status/worker/" | json_pp
{
"queued" : {
"default" : 0,
"high" : 825,
"low" : 655
}
}
Get crawling stats
$ curl -s "http://localhost:8000/api/stats/" | json_pp
{
"status" : [
{
"count" : 525,
"percentage" : 98.3,
"label" : "ok"
},
{
"label" : "error",
"percentage" : 1.3,
"count" : 7
},
{
"label" : "timeout",
"percentage" : 0.4,
"count" : 2
}
],
"status_codes" : [
{
"code" : 200,
"count" : 413,
"percentage" : 78.7
},
{
"code" : 501,
"percentage" : 12.4,
"count" : 65
},
{
"percentage" : 6.1,
"count" : 32,
"code" : 404
},
{
"code" : 500,
"percentage" : 2.7,
"count" : 14
},
{
"code" : 502,
"count" : 1,
"percentage" : 0.2
}
]
}
Using Webhook integration
** Set the config values**
Create a config.toml where your service and commands are launched, or specify a path to a TOML file via the HYDRA_SETTINGS environment variable. config.toml or equivalent will override values from udata_hydra/config_default.toml, lookup there for values that can/need to be defined.
UDATA_URI = "https://dev.local:7000/api/2"
UDATA_URI_API_KEY = "example.api.key"
SENTRY_DSN = "https://{my-sentry-dsn}"
The webhook integration sends HTTP messages to udata when resources are analyzed or checked to fill resources extras.
Regarding analysis, there is a phase called "change detection". It will try to guess if a resource has been modified based on different criterions:
- harvest modified date in catalog
- content-length and last-modified headers
- checksum comparison over time
The payload should look something like:
{
"analysis:filesize": 91661,
"analysis:mime-type": "application/zip",
"analysis:checksum": "bef1de04601dedaf2d127418759b16915ba083be",
"analysis:last-modified-at": "2022-11-27T23:00:54.762000",
"analysis:last-modified-detection": "harvest-resource-metadata",
}
Development
docker-compose
Multiple docker-compose files are provided:
- a minimal
docker-compose.ymlwith PostgreSQL docker-compose.broker.ymladds a Redis brokerdocker-compose.test.ymllaunches a test DB, needed to run tests
NB: you can launch compose from multiple files like this: docker-compose -f docker-compose.yml -f docker-compose.test.yml up
Logging & Debugging
The log level can be adjusted using the environment variable LOG_LEVEL.
For example, to set the log level to DEBUG when initializing the database, use LOG_LEVEL="DEBUG" udata-hydra init_db .
Writing a migration
- Add a file named
migrations/{YYYYMMDD}_{from}_up_{to}.sqland write the SQL you need to perform migration.fromshould be the revision from before (egrev1),tothe revision you're aiming at (egrev2) - Modify the latest revision (eg
rev2) inmigrations/_LATEST_REVISION udata-hydra migratewill use the info from_LATEST_REVISIONto upgrade torev2. You can also specifyudata-hydra migrate --revision rev2
Deployment
3 services need to be deployed for the full stack to run:
- worker
- api / app
- crawler
Refer to each section to learn how to launch them. The only differences from dev to prod are:
- use
HYDRA_SETTINGSenv var to point to your customconfig.toml - use
HYDRA_APP_SOCKET_PATHto configure where aiohttp should listen to a reverse proxy connection (eg nginx) and useudata-hydra-appto launch the app server
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for udata_hydra-1.0.1-py3-none-any.whl
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | a7219d9536c7e892471d66ad61b4433462f3bad8b97061bdc994cb224e3938da |
|
| MD5 | ce5f5db0e85c815a42f89edf3ed846bc |
|
| BLAKE2b-256 | c6c21349571f7288d79e7f3ae72ee9e820cf30407b3005b152e6b8d10f3b652c |
