I have been told it is bad practice to use Linux kill command to stop a StreamBase server instance. Is this correct? What should we do instead?

Note that MS Windows, "Task Manager", End Task function is analogous to Linux kill.

Platform administration guidance

StreamBase runs in a Java VM and should be stopped by using the commands:
  StreamBase 7: sbadmin shutdown
  StreamBase 10: epadmin servicename=node.cluster stop node

This allows all files to be flushed and closed and all connections to external systems to be ended with a good state. Adapters respond to the shutdown and stop commands by running their shutdown() callback method which allows them to clean up any state managed by the 3rd-party adapter API.

Any other means of stopping a StreamBase server risks data loss from missed writes to disk files and interrupted adapter connections.

For StreamBase 7, a sudden stop can damage disk-based query tables when using transaction mode 0, and can lose data when using transaction mode 1 (the default).

For StreamBase 10, a sudden stop will damage the transactional shared memory for this node, and the node will need to be reinstalled before it is restarted. However this does not result in data-loss for the cluster as long as any other node in the cluster is using the same partition for replication, and remains running or is shut down normally.