Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Use of a cluster file system for the Standalone Agent instance's journal is an option. However, this also brings performance penalties and requires user intervention to restart a failed Standalone Agent from a clean copy of its journal.
  • If transaction data for a Standalone Agent is lost then this will affect the state of orders currently running, until their execution status has been reported back to the Controller. When a Standalone Agent is started with a new journal then the Controller will automatically forward workflows and orders to bring the Agent's journal in sync with information from the Controller's journal.
  • Do not try to restore a version of the Agent's journal that is not in sync with the Controller. If an older version of an Agent's journal is restored then this will be inconsistent with latest changes to the Controller's journal and will prevent the Agent from being coupled. If you have to restore an Agent, for example after disk failure, then you have to accept that currently running orders are lost and you should start the Agent with a new journal. To this purpose you can remove the contents of the Agent's state directory.

Agent Cluster

  • The JS7 - Agent Cluster guarantees redundancy and consistency of transaction data which is synchronized between the Active and Standby Director Agent instances on different machines. If the Active Director Agent's journal is lost, for example due to disk failure, then the Standby Director Agent will pick up operation from a synchronized copy of the journal. If the failed Director Agent instance is started later on then it will be assigned the standby role and will synchronize its journal from the Active Director Agent.
  • Subagents do not use a journal as they are used for job execution only. Workflows and order state transitions are managed by the Director Agent only.

...