Add a no_overlay option for the rkt task config.

[kmalec]

Directly relates to rkt/rkt#1922 - for various reasons, I have to keep up with RHEL 7 systems, and --no-overlay for rkt run seems to help with the bug above.

[schmichael]

Have you built Nomad and tested this feature on an affected system?

[schmichael]

Backticks around the command line flag:

... use `--no-overlay=true` flag for ...

[kmalec]

The build I actually have running, but checking some test jobs from yesterday, I see a few alloc/UUID/task-name/secrets leftover tmpfs mounts in one of the systems.
rkt gc runs fine now, so --no-overlay hopefully helps, but I'm not sure where those secrets mounts come from. My environment is a bit volatile right now, maybe unrelated to this change.

[schmichael]

Fixed conflicts and am waiting on Travis.

[kmalec]

Okay, I did some more basic tests with a simple centos7 ACI on RHEL7.3 host.

• the no_overlay option resolves the GC issue referenced in the commit - for the price of performance hit of not using overlayfs, I can now rkt gc without those Device or resource busy on stage2 rootfs
• the leaking mountpoints of Nomad (tmpfs tmpfs 1.0M 0 1.0M 0% /data/nomad-data/alloc/e81f7fec-0d56-5970-6af2-3019343201b9/linuk-task/secrets) don't seem to be caused by this commit, as they happen with 0.5.6 client agent too. Not sure if that's tracked yet.

Was just worried this PR could be the cause of those mountpoints, but no :)

[schmichael]

Closing in favor of #2702 -- I had to fix a merge conflict and needed to push up the changes. Sorry! Will get #2702 merged as soon as Travis passes.

[github-actions]

I'm going to lock this pull request because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active contributions.
If you have found a problem that seems related to this change, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.