Bug #19429
Updated by Peter Amstutz almost 2 years ago
A failure mode with cloud compute nodes is that the scratch space can fail to mount due to misconfiguration. When this happens, jobs may mysteriously run out of disk space and it can be very difficult to debug.
Maybe "arvados-server cloudtest" could include a check that an expected amount of disk space is available, or even check on AWS that we have the right IAM permissions.