Sometimes the workspace for a job can get irrevocably corrupted (eg. the workspace is kept between jobs, and some cached data is incorrect because one of the executions was buggy). In these circumstances, the easiest way to resolve the problem is to delete the workspace and re-create it from scratch when the job next runs.
However, there is no obvious way to do this with Bamboo. This question is along the same lines, but ultimately did not find a satisfactory solution (especially when you are using multiple agents). To summarise, the options are:
- Temporarily modify the job definition to clean the workspace (eg. "Force clean build" in the source checkout) and run the job.
- Have a clean task permanently defined in your job definition, which is enabled with a variable. When you need to clean, then run the job with the cusotm variable set.
- Go through all the workspaces for all your agents on all your machines and manually delete the job's workspace from the file system.
Any of these optionas are "good enough" workarounds if you only have 1 agent. However, the first 2 don't work reliably when you have multiple agents (the workspace for only 1 of the agents will have been cleaned), and the final one doesn't scale well.
Are there any other approaches, or should I create a feature request?
