Some workers can use quite lots of disk space and memory for their computations (e.g. swh-loader-svn).

This can have quite some unfortunate side-effects:

1. Out of memory error -> worker's killed before it can actually empty its occupied disk

2. The following workers have less disk space to work, so they can then get stopped too early

3. Thus, this can cause filling up disk-space in the worker machine quite faster

4. logging more rapidly errors which can cause filling up disk in other machines as well (the ones centralizing logs).

5. workers consuming queues without really doing anything (which lead to losing sight on what is actually done)

Having some ways of :

• mitigating the memory occupation

• monitoring and cleaning up the useless dangling temporary directories

could help mitigate this.