Changes between Version 17 and Version 18 of Workshops/JobCheckpointing/Examples

Timestamp:

04/17/2026 03:59:04 PM (3 months ago)

Author:

Carl Baribault

Comment:

Added common application settings, "Make copies..." subtitle

Workshops/JobCheckpointing/Examples

@@ -998 +998 @@
 === Common features among the examples ===
 
-Each of the example application scripts included here are designed for use with the pair of job scripts [wiki://Workshops/JobCheckpointing/Examples#checkpoint_submitter.sh checkpoint_submitter.sh] and [wiki://Workshops/JobCheckpointing/Examples#checkpoint_runner.sh checkpoint_runner.sh]
+Each of the example application scripts included here are designed for use with the pair of job scripts.[wiki://Workshops/JobCheckpointing/Examples#checkpoint_submitter.sh checkpoint_submitter.sh] and [wiki://Workshops/JobCheckpointing/Examples#checkpoint_runner.sh checkpoint_runner.sh]
 
 As part of that design, each of the application scripts implements the following list of features or capabilities in its own specific language.
 
- 1. Each application script handles the SIGTERM signal sent by the [wiki://Workshops/JobCheckpointing/Examples#CheckpointRunner Checkpoint Runner] job immediately before that job times out.
- 2. Each application script performs its own periodic '''checkpointing''' or saving intermediate or partial results, periodically.
- 3. Each application script resumes from the most recently recorded checkpoint, if any, when invoked by the [wiki://Workshops/JobCheckpointing/Examples#CheckpointRunner Checkpoint Runner].
-
-=== Make copies of the BASH scripts checkpoint_submitter.sh and checkpoint_runner.sh
+ 1. Each application script handles the '''SIGTERM signal''' sent by the [wiki://Workshops/JobCheckpointing/Examples#CheckpointRunner Checkpoint Runner] job immediately before that job times out.
+ 2. Each application script performs its own '''periodic checkpointing''' or saving intermediate or partial results, periodically.
+ 3. Each application script '''resumes''' from the most recently saved checkpoint, if any, when invoked by the [wiki://Workshops/JobCheckpointing/Examples#CheckpointRunner Checkpoint Runner].
+ 4. Each application script uses the following settings in the form of '''BASH environment variables''' (see  [https://wiki.hpc.tulane.edu/trac/wiki/cypress/WorkingOnClusters#EnvironmentVariables Environment Variables]) with default values that can be overridden beforehand or at time of submittal of the [wiki://Workshops/JobCheckpointing/Examples#checkpoint_submitter.sh checkpoint_submitter.sh] job script. (See [wiki://Workshops/JobCheckpointing/Examples#Overridingdefaultsettings Overriding default settings].)
+  * '''CKPT_PATH''' - file path of checkpoint value (default varies)
+  * '''CHECKPOINT_EVERY''' - number of loop iterations between periodic checkpoints (defaults to 20)
+  * '''MAX_ITER''' - maximum number of loop iterations to perform (defaults to 500)                                                                
+
+=== Make copies of the BASH and Python scripts ===
+
+To use any of the checkpointing examples mentioned below, you'll need to make your own copies of the required BASH and Python scripts in your own environment on Cypress.
+
+==== Make copies of the BASH scripts checkpoint_submitter.sh and checkpoint_runner.sh ====
 
 All of the examples linked further below make use of the following two (2) BASH job scripts mentioned above.
@@ -1013 +1021 @@
 * [wiki://Workshops/JobCheckpointing/Examples#checkpoint_runner.sh checkpoint_runner.sh] 
 
-=== Make copies of the Python scripts aggregate_usage.py and plot_progress.py
+==== Make copies of the Python scripts aggregate_usage.py and plot_progress.py ====
 
 Also, the '''checkpoint_submitter.sh''' script calls out the following two (2) Python scripts for post-workflow analysis of execution time and memory usage by the checkpoint runner jobs.