Troubleshooting

Introduction

Problems regarding AdaptiveFlow and approaches to solving them can be classified into two categories:

1. General problems, where the precise problem is yet to be determined.

2. Specific problems, where the precise problem could already be determined.

Accordingly, we classify the approaches for troubleshooting into these two categories:

1. General problems/approaches

2. Specific problems/approaches

General Problems/Approaches

Overview of general approaches:

• Increasing the command-verbosity

• Inspecting log files and workflow files

  • Logfiles in workflow/output-files/jobs/

  • Logfiles in workflow/output-files/queues/

  • Ligand processing status logs in workflow/ligand-collections/ligand-lists/

• Interactive jobs

Increasing the Command-Verbosity

The verbosity of the AdaptiveFlow end-user commands (starting with aflp_ or afvs_) can be increased by setting the option verbosity_commands in the control file to the value 'debug':

verbosity_commands=debug

Inspecting log files and workflow files

The verbosity of the AdaptiveFlow logfiles (in the workflow/output-files/ folders) be increased by setting the option verbosity_logfiles in the control file to the value 'debug':

verbosity_logfiles=debug

This will affect the following log files:

• Logfiles in workflow/output-files/jobs/

• Logfiles in workflow/output-files/queues/

In addition, the processing status of each ligand and possible errors related to the processing steps can be found int the the subfolders of the workflow/ligand-collections/ligand-lists/ folder.

Interactive Jobs

Interactive jobs can for instance be useful when AdaptiveFlow starts the workflow and begins using the local storage of the compute nodes, but then crashes and does not manage to copy back the log-files which are stored on the compute nodes. If this case, if one is running AdaptiveFlow via an interactive job and is logged in to the compute node, one has immediate access to all the workflow files of AdaptiveFlow on that compute node.

Similarly, one can run non-interactive jobs and then ssh into the the compute node which is used by AdaptiveFlow. However, this does not always work because if the job crashes and terminates, ssh access is often automatically terminated as well to that compute node (since one may not have any more job running on that node, which is often a requirement for ssh access to a compute node).

Specific Problems/Approaches

[Will be added with feedback from future users.]