Working with Enterprise Search Part 2

This is a live blogging post from the International SharePoint Conference London 2012. So don’t expect well-written proza here.

Session by Neil Hodgkinson

Service Message: A lot of demo’s in this session, so not much content in this post

Scalability

  • designing for size
  • know what to scale out to obtain which result
  • the only way to change search in powershell is to clone and change existing topologies

Crawler scale out

  • Failover between crawlers will occur after 60 mins without updates
  • Failed crawler will recover automatically

Query scale out

  • Failover between query processors will occur immediately
  • Failover partition don’t take part in the query at all normally, only at failover
  • When a complete partition is gone, you might still get a partial resultset if the disk is still readable
  • Having 3 components is better than 2 because there must be 2 responding components to initiate auto recovery

Going Deeper

  • Repartitioning process is a complex beast : )
  • host distribution rules: allocate start address to crawl databases, solves the slow host problem
  • host distribution rules are a supported way to truncate your crawl tables