Crawler for news based on StormCrawler. Produces WARC files to be stored as part of the Common Crawl. The data is hosted as AWS Open Data Set – if you want to use the data and not the crawler software please read the announcement of the news dataset.



The default configuration should work out-of-the-box. The only thing to do is to configure the user agent properties send in the HTTP request header. Open the file conf/crawler-conf.yaml in an editor and fill in the values for and all further properties starting with the http.agent. prefix.

Run the crawl

Generate an uberjar:

mvn clean package

And run ...

storm jar target/crawler-1.16.jar -conf $PWD/conf/es-conf.yaml -conf $PWD/conf/crawler-conf.yaml $PWD/seeds/ feeds.txt

This will launch the crawl topology. It will also "inject" all URLs found in the file ./seeds/feeds.txt in the status index. The URLs point to news feeds and sitemaps from which links to news articles are extracted and fetched. The topology will create WARC files in the directory specified in the configuration under the key warc.dir. This directory must be created beforehand.

Of course, it's also possible to add (or remove) the seeds (feeds and sitemaps) using the Elasticsearch API. In this case, the can topology can be run without the last two arguments.

Alternatively, the topology can be run from the crawler.flux, please see the Storm Flux documentation. Make sure to adapt the Flux definition to your needs!

Monitor the crawl

When the topology is running you can check that URLs have been injected and news are getting fetched on [http://localhost:9200/status/_search?pretty]. Or use StormCrawler's Kibana dashboards to monitor the crawling process. See instructions on [] to install the templates for Kibana.

There is also a shell script bin/es_status to get aggregated counts from the status index, and to add, delete or force a re-fetch of URLs. E.g.,

$> bin/es_status aggregate_status
34      FETCHED

Run Crawl from Docker Container

First, download Apache Storm 1.2.3. from the download page and place it in the directory downloads:

mkdir downloads
wget -q -P downloads --timestamping$STORM_VERSION/apache-storm-$STORM_VERSION.tar.gz

Then build the Docker image from the Dockerfile:

docker build -t newscrawler:1.16 .

Note: the uberjar is included in the Docker image and needs to be built first (see above).

Launch an interactive container:

docker run --net=host \
    -p \
    -p 5601:5601 -p 8080:8080 \
    -v .../newscrawl/elasticsearch:/data/elasticsearch \
    -v .../newscrawl/warc:/data/warc \
    --rm -i -t newscrawler:1.16 /bin/bash

NOTE: don't forget to adapt the paths to mounted volumes used to persist data on the host.

CAVEAT: Make sure that the Elasticsearch port 9200 is not already in use or mapped by a running ES instance. Otherwise ES commands may affect the running instance!

The crawler is launched in the running container by the script


After 1-2 minutes if everything is up, connect to Elasticsearch on port 9200 or Kibana on port 5601.