How Recovery Works

During execution of a process flow, at every checkpoint BPM Server stores the current state of the process flow in a recovery file. There is one recovery file for each execution of a process flow. These recovery files are stored in a recovery folder defined in the property abpm.recovery.repository.root. If kernel goes down during execution of process flow, recovery file will have the state of the process flow till the last checkpoint. If no checkpoint is reached, no recovery file is created and the process flow can never be recovered after failure. In the recovery file only state of the process flow is saved. Intermediate data of the process flows are stored in the repository folder. Once system is restarted and if recovery is enabled, BPM Server looks for the state of the uncompleted process flow in the recovery file, picks up the intermediate data from the repository file and resumes the process flow.

The recovery file, remains in the recovery folder unless the process flow execution is completed. The recovery of process flows cannot be completed if intermediate data are not available in repository folder. Intermediate data gets deleted automatically by Data Cleanup or user can delete it manually to free disk space. To know more about data cleanup, refer to the section Data Cleanup.
Even if recovery is not enabled, the recovery files created during execution of the process flows are stored in the recovery folder. In this case recovery of the uncompleted process flows are not done. Later on if you want to recover those process flows, you need to enable recovery and restart the kernel. After Kernel is restarted, the uncompleted process flows are recovered.