Evolving to Open: Adopting Apache Solr for the Red Hat Customer Portal: Presented by Chris Bredesen, Red Hat

O C T O B E R 1 3 - 1 6 , 2 0 1 6 • A U S T I N , T X

Evolving to OpenAdopting Apache Solr for the Red Hat Customer Portal

Chris Bredesen

Manager, Software Engineering at Red Hat

@cbredesen

01

Hello!

$ rm -rf agenda$

About Red Hat

● Largest open source company● Subscription model● Communities matter

Red Hat Customer Portal

● Apps, services & content● Delivers subscriber value● On-site search is critical

Search Infrastructure 1.0

Search Infrastructure 1.1

● More content from new sources● Attempts to optimize crawling using pushed feeds

● Desire for index-based browse UI

Growing Pains

● 150K+ documents w/per-document licensing

● Want to add 10M+ more :-o● OK for turn-key; lacks customizability● Not suitable for browse● Click-through not stellar● Proprietary “black box”

Lucene

● Highly tuned recommendation engine● Narrow focus, deep integration● Back-testable against all cases/solutions

● Success!

Search Infrastructure 2.0

Challenges

● Web search successful● Internal search failed – crash and burn on tokens like CVE-2001-001

● Underestimated indexer dev effort● Indexers misbehaving has larger impact

● Solr Java issues @ 10X load

Results

● Click through rate sustained @ 2X● Query response consistently sub-second (including our auth layer)

● Developers thrilled with API2.0 Release

Future Plans

● Replace some custom indexer code with Fusion● Signals API for active tuning● Admin tooling for tuning, elevation, synonyms/spell-check

Discussion

Q&A