Kb130

Emulab FAQ: Testbed Operations: What history is kept for failed experiments?

When an experiment is first started (created), its history is removed if it fails immediately. Its only when the swap interface is used that all the history is saved.

In the start/fail case I maintain a single backup directory for debugging. /usr/testbed/expwork/pid/eid-failed. I did this to prevent cluttering up the archive with stuff that does not work, since the vast majority of start errors are parse errors, missing rpm/tarball errors, bad OSID errors, etc.

On the rare occasions that its an interesting error, we have the -failed directory to go to. I honestly do not think it is necessary to archive failed experiments.