Mozilla Services Data Pipeline

This repository contains the extra bits and pieces needed to build heka for use in the Cloud Services Data Pipeline.

Visit us on irc.mozilla.org in #datapipeline.

Building a Data Pipeline RPM

Run bash bin/build_pipeline_heka.sh from the top level of this repo to build a heka RPM.

Using the Data Pipeline

If you are simply looking to test out some data analysis plugins and don't want to setup your own pipeline here is the fastest way to get going: https://mana.mozilla.org/wiki/display/CLOUDSERVICES/Using+the+sandbox+manager+in+the+prod+prototype+pipeline

Running/Testing Your Own Data Pipeline

You can set up a bare-bones data pipeline of your own. You will get an endpoint that listens for HTTP POST requests, performs GeoIP lookups, and wraps them up in protobuf messages. These messages will be relayed to a stream-processor, and will be output to a local store on disk. There will be basic web-based monitoring, and the ability to add your own stream processing filters.

1. Clone this data-pipeline github repo

   git clone https://github.com/mozilla-services/data-pipeline.git
   

2. Build and configure heka. If you are unable to build heka, drop by #datapipeline on irc.mozilla.org and we will try to provide you a pre-built version.

3. Make sure you have the depencies installed: 1. OpenSSL v1.0+ (required by lua_openssl) 2. libpq, the PostgreSQL API

4. Run bash bin/build_pipeline_heka.sh

5. Install lua modules

   mkdir lua_modules
   rsync -av build/heka/build/heka/lib/luasandbox/modules/ lua_modules/
   rsync -av hindsight/modules/ hindsight/io_modules/ lua_modules/
   

6. Procure a GeoLiteCity.dat file and put it in the current dir

   wget http://geolite.maxmind.com/download/geoip/database/GeoLiteCity.dat.gz
   gunzip GeoLiteCity.dat.gz
   

7. Set up the main Pipeline using the examples/basic_local_pipeline.toml config file. This will listen for HTTP POSTs on port 8080, log the raw and decoded messages requests to stdout, run the example filter, and output the records to a file.

   build/heka/build/heka/bin/hekad -config examples/basic_local_pipeline.toml
   

8. Check the monitoring dashboard at http://localhost:4352

9. Fire off some test submissions!

   for f in $(seq 1 20); do
     curl -X POST "http://localhost:8080/submit/test/$f/foo/bar/baz" -d "{\"test\":$f}"
   done
   

10. Verify that your data was stored in the output file using the heka-cat utility

    build/heka/build/heka/bin/heka-cat data_raw.out
    build/heka/build/heka/bin/heka-cat data_decoded.out
    

11. Experiment with sandbox filters, outputs, and configurations.

Useful things to know

• GeoIP
  • It’s not terribly interesting to do GeoIP lookups on 127.0.0.1, so you may want to provide a --header "X-Forwarded-For: 8.8.8.8" argument to your curl commands. That will force a geoIP lookup on the specified IP address (Google’s DNS server in this example).
• How to configure namespaces
  • The example config allows submissions to either /submit/telemetry/docid/more/path/stuff or /submit/test/id/and/so/on
  • You can add more endpoints by modifying the namespace_config parameter in basic_local_pipeline.edge.toml.
  • The namespace config is more manageable if you the JSON in a separate file, and run it through something like jq -c '.' < my_namespaces.json before putting it into the toml config.
• Where to get more info about configuring heka
  • http://hekad.readthedocs.org/en/latest/index.html