README.md

In PRL-PRG/rapr: Run R experiments

Running scripts on each package

Job failures

Individual jobs can fail. The scripts should be done in a way that if a job fail, the exit code of the script should be non-zero. This way, it will be recorded in the GNU parallel log file and can be easily retried.

To retry failed jobs, run the on-each-package.sh with PARALLEL_ARGS="--retry-failed" environment variable:

PARALLEL_ARGS="--retry-failed" ./runr/inst/on-each-package.sh ./runr/inst/tasks/package-coverage.R

This will rerun all jobs whose Exitval in parallel.log is non-zero. The log fill will have all new jobs run appended, therefore there will be duplicates and the log file should be normalized by running:

./runr/inst/normalize-parallel-log.R run/package-coverage/parallel.log

which will only keep the latest entry for each job.

Timeout

The default timeout is 30min. It can be changed by the TIMEOUT environment variable:

TIMEOUT=1m ./runr/inst/on-each-package.sh ./runr/inst/tasks/package-load.R

PRL-PRG/rapr documentation built on Oct. 18, 2022, 5:45 a.m.