Having Your Cake and Eating It Too With Apache OODT and Apache Solr Andrew F. Hart Paul M. Ramirez About Myself… • Software Engineer – NASA Jet Propulsion Laboratory – “Data Management” • Committer: – OODT, SIS, Gora, Streams (Incubating) • Mentor: Streams (Incubating) What We’ll Cover • Overview of OODT & Solr Projects • Strategies for Combining OODT and Solr • Detailed Deployment/Config. Example • Where to Learn More & Participate Apache OODT • Object Oriented Data Technology • Origin in NASA mission data systems • Components for – Information integration – Data cataloging and archiving – Configurable workflow processing Apache OODT • OODT @ Apache – Incubation: 2010, Graduation: 2011 – 29 Committers – Latest Release: 0.5 (Dec. 26, 2012) Apache OODT • Karoo Array Telescope (KAT-7) Apache OODT • Virtual Pediatric Intensive Care Unit Apache OODT • Regional Climate Model Evaluation System Apache OODT • Commonalities between systems – Lots of data – Defined processing steps / algorithms • Archives important (… search important) Apache OODT • Strengths of OODT for the above use cases – Loosely coupled components – Standard protocols, well-defined interfaces – Highly configurable – Vetted, reliable code Apache Solr • Search + Web Services – Powerful features – Flexible formats – Highly configurable Apache Solr • The White House Apache Solr • Netflix Apache Solr • NASA Planetary Data System OODT & Solr • Why use these projects together? • Archives often need search capability • Similarities / Compatibilities – XML-based configuration – Environment (Java, Tomcat) Example Integration “Standard” Data Archive Pipeline Example Integration “Standard” Data Archive Pipeline + Search OODT Products • Typically 1-1 with Files • Each uniquely identifiable (GUID) • Support for higher-level “ProductType” – A way to define collections OODT Metadata • • • • Annotations for products Key:{Val|Multival} Common across all OODT components Two general classes: – System – User OODT Metadata • System Metadata – Added automatically by OODT Components – Used to track state – Used to encode relationships between data OODT Metadata • User Metadata – Specified as “policy” – Can be product-level, or productTypelevel – Used to extract & persist information from files as they are ingested (become products) OODT Metadata • Metadata (Policy) Example (external) Solr Schema • XML document • Define what will be indexed (“Fields”) • Provide high-level context hints – Data type, behavior, pre-processing • Extremely flexible, extensible Solr Schema • Solr Schema Example (external) Making the Connection • SolrIndexer Tool – Part of the File Manager component tools – Map OODT Metadata to Solr Fields – Create Solr documents from OODT products – Note: only talking about metadata SolrIndexer Tool • Org.Apache.Oodt.Cas.Filemgr.Tools • Available since 0.4 Release • Recommend to use 0.5+ as some stability improvements were added • Several modes of operation SolrIndexer Tool SolrIndexerTool • Invocation Examples: Ingest all products from the specified File Manager instance java -DSOLR_INDEXER_CONFIG=/path/to/indexer.properties -Djava.ext.dirs=/path/to/cas/filemgr/lib/ \ org.apache.oodt.cas.filemgr.tools.SolrIndexer \ --all \ --fmUrl http://localhost:9000 \ --solrUrl http://localhost:8080/solr \ SolrIndexerTool • Invocation Examples: Ingest all products from the specified ProductType(s) java -DSOLR_INDEXER_CONFIG=/path/to/indexer.properties -Djava.ext.dirs=/path/to/cas/filemgr/lib/ \ org.apache.oodt.cas.filemgr.tools.SolrIndexer \ --types urn:some:ProductType \ --fmUrl http://localhost:9000 \ --solrUrl http://localhost:8080/solr \ SolrIndexerTool • Invocation Examples: Ingest a single product by its unique product id java -DSOLR_INDEXER_CONFIG=/path/to/indexer.properties \ -Djava.ext.dirs=/path/to/cas/filemgr/lib/ \ org.apache.oodt.cas.filemgr.tools.SolrIndexer \ --product 19bcb4b8-7999-11e1-b581-8b771498975d \ [--delete] \ --fmUrl http://localhost:9000 \ --solrUrl http://localhost:8080/solr SolrIndexerTool • Invocation Examples: Force optimization of the Solr index java -DSOLR_INDEXER_CONFIG=/path/to/indexer.properties \ -Djava.ext.dirs=/path/to/cas/filemgr/lib/ \ org.apache.oodt.cas.filemgr.tools.SolrIndexer \ --optimize --solrUrl http://localhost:8080/solr Indexer.properties • Configuration file for the SolrIndexer • Specify mapping between OODT product metadata and Solr fields • Additional “pre-processing” features Indexer.properties • Example Indexer.properties file (external) Use Case I • • • • Building a searchable data archive “Long-term” / “Lights-out” archive Products & metadata immutable Many NASA mission data systems use this model • Want to make it easily searchable Use Case I “Standard” Data Archive Pipeline + Search Use Cases II • Building an interactively editable, searchable data archive • Data and metadata mutable • Want to dynamically select product(s) to edit based on metadata Use Case II Interactively Editable Data Archive Pipeline + Search Use Case II Interactively Editable Data Archive Pipeline + Search Solr catalog out of sync! Synchronization • Two ways (at least) to solve this: A. Modify the OODT Curator Services B. Treat OODT Curator Services as “black box” and write “wrapper” service to invoke Curator Services AND update Solr (via scripted call to SolrIndexer, for example) Modify Curator Services • Services implemented in JAX-RS • /curator/src/main/java/org/apache/oodt/cas/cu ration/service • [curator_url]/services/metadata/update • Options: – Utilize Solr Java API – Wrap call to OODT SolrIndexer tool Use Case II-A Modified Curator Services to Simultaneously update Solr Example • Interactive event tagging Wrap Curator Services • Curator Service/API is “black box” • Develop custom service that: – Issues POST request to Curator service – Updates Solr index via, e.g.: • Utilize Solr Java API • Wrap call to OODT SolrIndexer tool Use Case II-B Wrapping OODT Curation Services with Custom UI & Services Example Lessons • Solr compliments OODT File Manager • RESTful interfaces (Solr + OODT Curator) allow for great flexibility in designing services and UI • “Best” approach depends on situation Next Steps • Develop “SolrCatalog” for OODT File Manager? – Pros: Reduction in “moving parts” – Cons: Restrictive? • Implement Use Case II-A as optional mode for Curator web service layer Learning More • Solr – http://lucene.apache.org/solr • solr-user@lucene.apache.org • OODT – http://oodt.apache.org • https://cwiki.apache.org/confluence/display/O ODT/Home • oodt-user@apache.org Thanks! • Questions?