Fork me on GitHub


  • Nirmata Workflow uses Apache ZooKeeper which has certain built-in limits. For example, the entire ZooKeeper database is kept in memory. So, huge data sets are not supported.
  • A given Nirmata Workflow run should be able to support several thousand tasks.
  • Each task should limit the amount of result data it produces. Result data is meant for a few dozen values at most.
  • It's vital that you clean up old runs periodically. See the Administration APIs for details.
  • For proper high availability and failover, you should run a cluster of ZooKeeper instances (as opposed to just one).