Indexing a File Containing Tweets with Flume SpoolDirectorySource

SpoolDirectorySource specifies a directory on a local disk that Flume monitors. Flume automatically transfers data from files in this directory to Solr. SpoolDirectorySource sends data using a channel to a sink, in this case a SolrSink.

The tutorial provides examples that work with an environment established using a package-based installation. If you installed Cloudera Search using parcels, adjust file paths accordingly.

1. Delete all existing documents in Solr:

   $ sudo /etc/init.d/flume-ng-agent stop
   $ solrctl collection --deletedocs collection3

2. Comment out TwitterSource and HTTPSource in /etc/flume-ng/conf/flume.conf and uncomment SpoolDirectorySource:

   # Comment out “agent.sources = httpSrc”
   # uncomment “agent.sources = spoolSrc”

3. Delete any old spool directory and create a new spool directory:

   $ rm -fr /tmp/myspooldir
   $ sudo -u flume mkdir /tmp/myspooldir

4. Restart the Flume Agent:

   $ sudo /etc/init.d/flume-ng-agent restart

5. Send a file containing tweets to the SpoolDirectorySource. Usethe copy-then-atomic-move file system trickto ensure no partial files are ingested:

   $ sudo -u flume cp \
   /usr/share/doc/search*/examples/test-documents/sample-statuses-20120906-141433-medium.avro \
   /tmp/myspooldir/.sample-statuses-20120906-141433-medium.avro
   $ sudo -u flume mv /tmp/myspooldir/.sample-statuses-20120906-141433-medium.avro \
   /tmp/myspooldir/sample-statuses-20120906-141433-medium.avro

6. Check the log for status or errors.

   $ cat /var/log/flume-ng/flume.log

7. Check the completion status.

   $ find /tmp/myspooldir

Use the Cloudera Search GUI. For example, for the localhost, use http://localhost:8983/solr/collection3/select?q=*%3A*&wt=json&indent=true to verify that new tweets have been ingested into Solr.