Computing grids consist of a large-scale, highly-distributed hardware architecture, often built in a hierarchical way, as cluster federations. At such scales, failures are no longer exceptions, but part of the normal behavior. When designing software for grids, developers have to take failures into account, in order to be able to provide a stable service. The fault-tolerance mechanisms need to be validated and evaluated. It is therefore crucial to make experiments at a large scale, with various volatility conditions, in order to measure the impact of failures on the whole system. This paper presents an experimental tool allowing the user to control the volatility conditions during a practical evaluation of fault-tolerant systems. The tool is based on failure-injection mechanisms. We illustrate the usefulness of our tool through an evaluation of a failure detector specifically designed for hierarchical grids.