MountFlags "slave" helps to prevent "device busy" errors on RHEL/CentOS 7.3 kernels

[jgagnon1]

MountFlags "slave" helps to prevent "device busy" errors on RHEL/CentOS 7.3 kernels

moby/moby#34198

[haxorof]

@jgagnon1 Thanks for the PR! I will have a look at it later today.

[haxorof]

@jgagnon1 I have now looked at your contribution and I think it is it good in general. What I am thinking about it the variable docker_setup_mountslaveflag which I guess you set manually if you bump into this issue or not? To my understanding this is an issue in the older kernels like 3.10 which is installed in CentOS 7.3. Have you checked if this can generally be applied to CentOS 7.3 or not?

I will have some more time today to read through the issues related the moby issue you linked.

Cheers!

[jgagnon1]

Yes, since I have an heterogeneous environment - I just have host group with that variable set to true, this is optional since this might not be an issue on 4.x kernels.

I am willing to document that flag to README file if you think this would be helpful (I do).

[haxorof]

What I am actually thinking about if this is something you always want to set when you are running a Linux kernel v3 now. What do you think about that? In your case it is like that right? I know I have seen this issue on CentOS 7.3 also but did not know it was related to the mount flag then.

[jgagnon1]

I am not 100% sure this entirely solves the issue, but from my observations so far it seems to have solved it. I think this might be good if we want to enable it for all CentOS v3.x kernels.

[haxorof]

Sorry for my late answer here, been busy.

Maybe it would then be possible to replace the docker_setup_mountslaveflag with this kind of check {{ ansible_kernel|version_compare('4', '<') }}. What do you think about that?

[jgagnon1]

I am totally fine with it, pushed commit for it.

[haxorof]

Thanks @jgagnon1 for your contribution!