openark/orchestrator v2.1.4

GA release v2.1.4

on GitHub

Changes since last release: v2.1.2...v2.1.4

Notable:

• Important bugfix in recovery flow; some failure detections could go unhandled depending on concurrency behavior, see #196
• Added -c force-master-failover command, which just kicks in a failover for a cluster; there is no designated server to promote, orchestrator just makes believe master is dead and does whatever it does to fix the situation.
  • likewise /api/force-master-failover/:hostname/:port
• The Great Configuration Variables Exodus (deprecation) -- deprecating many variables; more work to come
• lost-in-recovery: masters who have been lost in recovery are:
  • downtimed forever
  • or until found to be healthy in a replication environment
• Concurrent probing of topology servers: reduces per-server discovery time by some 70% on high latency networks
• More visibility in failure detection process:
  • Better logging
  • Introduced "is actionable" flag for detection
  • Allows multiple detections in backend database for same instance/incident: up to one detection entry for "actionable", up to one detection entry for "not actionable". This gives some more (though not full) visibility into escalation process.
• added /api/leader-check: returns 200 when node is the leader, 404 if not.
• more... thanks @sjmudd for continued contributions