Yokozuna Release Notes
The third pre-release of Yokozuna.
Listen on canonical Solr URL (yz_39).
Added new target (
yz-setup) to Riak branch to make it easier to build Yokozuna/Riak from source (d32a20).
Disable Solr realtime get. It is not used by Yokozuna and greatly reduced write performance when enabled (9e7e4f).
Return a better error message when searching nonexistent index (yz_33).
Don't require a
content-typeheader when creating an index with no body (74492c).
Use Solr's JSON update instead of XML update. One benchmark showed approximately 15% improvement in write throughput (yz_31).
Pre-create the default schema (7a2167).
Add basic support for storing and retrieving of schemas via HTTP. Thanks to Marcel Neuhausler (yz_42).
Don't overwrite modified Solr Core config files (yz_14).
Update to the latest Riak.
This is the second pre-release of Yokozuna.
Active Anti-Entropy is a background process which constantly seeks out divergence between Riak data and indexes stored by Yokozuna. This process is made efficient by the use of hashtrees which are updated as data comes in. In the case were there is little to no divergence the work performed is unnoticeable. When divergence is detected read-repair is invoked and re-indexing occurs.
Eric Redmond added sibling support. Now, when allow_mult is enabled and siblings occur Yokozuna will index all object versions. Likewise, when siblings are resolved the obsolete indexes will be cleaned up.
Auto-Suffix JSON Extractor
Dan Reverri added a JSON extractor that automatically adds type suffixes to each field. This allows you to use the default schema with existing data. That is, you don't have to create a custom schema or add field suffixes yourself.
A collection of scripts have been created under
They are used for automating the process of running benchmarks with
Basho Bench, collecting various metrics, transforming the raw data,
and finally visualizing it via d3js. Currently these scripts assume
the target cluster is running on SmartOS. The driver scripts,
i.e. the scripts that run the benchmark, have only been tested on OSX.
Use semver rebar plugin (3bff0c).
Fix type issue in JSON extractor (f2d798).
Allow POST select (yz_23).
Send/Return Solr Headers (yz_29).
Remove shell wrapper around JVM (yz_18).
Remove unused files from included Solr distribution (ab1303).
This is the first pre-release of Yokozuna. It provides automatic distribution and replication of Solr indexes. It integrates with Riak KV providing near-real-time indexing of KV objects as they are written. Following are some of the major features included in this release.
Tight Integration With Solr
Yokozuna comes pre-bundled with Solr 4.0.0 running in the Jetty container. Yokozuna handles the basic administration of Solr such as starting and stopping Solr, creating Cores, sharding the documents, replicating the documents, and performing distributed queries.
To the client Yokozuna acts as a single Solr instance. Queries are sent to Yokozuna the same as a single Solr instance. Yokozuna determines which shards need to be contacted and performs a distributed Solr query. The results are returned verbatim to the client. This means if Solr distributed search supports it then Yokozuna supports it. Furthermore, existing Solr clients may be used to query Yokozuna. There is a simple example of using the SolrJ client.
Administration of Indexes via HTTP
To start indexing KV data an index must be created and a hook installed on the bucket. This can be done in one step via HTTP. Learn more at ADMIN.md.
Text, XML and JSON Extractors
Before Yokozuna can index a KV object it must first convert it to a Solr document. This release includes support for text, XML and JSON. To see how Yokozuna is extracting data the HTTP extract resource can be used for testing. It returns the field-value pairs in JSON format.
curl -XPUT -H 'content-type: application/json' 'http://localhost:8098/extract' -d '...SOME JSON...' | jsonpp
Tagging is the ability to add indexes via Riak KV metadata. This is useful when storing binary data or for indexing data about the data. More information can be found in TAGGING.md.
A Yokozuna AMI is provided to make it easier to try. More information can be be found in EC2.md.