CTDB: Where to from here and how can we get there?

WhoMartin Schwenke
WhereSambaXP 2015, Göttingen, Germany
WhenMay 2015
Abstract

CTDB needs to be cleanly restructured into some well defined components. The most obvious is a slimmer clustered TDB daemon, which is CTDB's main feature. The others components include separate daemons for managing:

  • cluster membership and leadership
  • public IP addresses
  • services

CTDB also needs well-defined protocols and a modern messaging system. Due to limited development resources it is unlikely that a shiny new version will be produced in a hurry. Therefore, this needs to be done step-wise, while maintaining a working version. How can we accomplish this? I'll talk about ideas like ctdb-tool, tunnelling a new protocol using the existing one, extensive use of callbacks and replicated internal TDBs. Hopefully this will start a conversation that, in the long term, will result in a better CTDB.

Slides
PDF (713 KB)