Need to check to see if a disk is mounted before using it.

[dvalinrh]

We are running into a situation where disks get renamed on the system (install usually), and when the scenario file is used it ends up overwriting the boot disk. We need to check to see if the device is in use. If so, abort the script with the appropriate message. The grab_disks util can be used to get a list of the devices that are currently not mounted.

[frival]

Note that grab_disks won't necessarily work for bare metal systems because there could be e.g. SAS disks meant for OS installs that aren't yet used but also shouldn't be used for a NVMe-based benchmark. We could add another feature to grab_disks to only take a certain type of disk but that also won't solve 100% of the issues as we have bare metal hosts that use e.g. M.2 NVMe drives for boot and those shouldn't be selected for benchmarking either.

TL;DR: I think we have to take whatever the user says to use (what we do now) but also "first do no harm" and not destroy any in-use disks. Wiping out the OS install really gives the whole ecosystem a bad look and is a really negative user experience. We could add a "--dude_trust_me" option to use drives that persist in reporting as in-use that actually aren't if we really wanted to but I'd like to see a use case before suggesting that.

[malucius-rh]

There's already a call to the "detect_mounts" tool, right after we check for "block device-ness," so we should be picking up anything currently mounted. It may make sense to add an "I see LVM partitions on the disk and they might be important" check.

[dvalinrh]

maybe use fdisk -l . If there is a Type associated with the partition, we abort. Allow an option to ignore that particular enforcement. Lets talk about it at the team meeting.