autoclave: a pressure cooker for programs
Summary
autoclave repeatedly executes a command line until its process exits with a non-zero status, is stopped/terminated by a signal, a user-specified timeout (-t) runs out, or a user-specified number of runs (-r) have passed without any failures.
It can log the contents of stdout (-l) and/or stderr (-e), and can rotate the log files so a fixed number of logs are kept (-c).
On failure, a handler program (-x) can be called with information about the child process. That way, a debugger can be opened if the stressed program times out or dumps core. For particularly long-running stress tests, the handler may also be a great place to send a notification.
For more detailed usage info, see the examples below the man page, and this blog post.
Examples
Repeatedly run buggy_program until it crashes:
$ autoclave buggy_program
Same, but print run and failure counts:
$ autoclave -v buggy_program
Same, but if it succeeds 10 times, return EXIT_SUCCESS:
$ autoclave -v -r 10 buggy_program
Same, but ensure there is at least 500 mesc between runs, sleeping if necessary:
$ autoclave -v -r 10 -m 500 buggy_program
Run without any delay:
$ autoclave -v -r 10 -m 0 buggy_program
Repeatedly run buggy_program, logging stdout to
buggy_program.1.stdout.log, buggy_program.2.stdout.log, and so on:
$ autoclave -l buggy_program
Same, but log to /tmp/buggy.ID.stdout.log instead:
$ autoclave -l -o /tmp/buggy buggy_program
Log, but only keep the 5 most recent log files:
$ autoclave -l -c 5 buggy_program
Same, but log stderr as well as stdout (keeping 10 files):
$ autoclave -l -e -c 5 buggy_program
A shortcut for autoclave -l -e -v ('s' for 'supervise'):
$ autoclave -s buggy_program
Repeatedly run buggy_program until it has failed 10 times:
$ autoclave -f 10 buggy_program
Run a program that occasionally deadlocks, halting it and counting it as a failure if it takes more than 10 seconds to complete:
$ autoclave -t 10 examples/deadlock_example
Same, but run the script examples/gdb_it when it times out, so gdb can
attach to the stopped process and investigate what is deadlocking.
(Note: the process is not halted after the -x command returns.)
$ autoclave -t 10 -x examples/gdb_it examples/deadlock_example
Run examples/crash_example, calling examples/gdb_it if it fails:
$ autoclave -x examples/gdb_it examples/crash_example