Docker container error 'Device not available' due to migrated z/OS mainframe datasets - how to overcome this?

[Craig Heron]

The ‘device not available’ error is causing problems starting the docker container as the NFS mount issued within the container references mainframe datasets of which the majority are in a migrated status (migrated off disk to tape).

These mainframe datasets are migrated off disk to avoid disk space issues and is standard practice.

We have NFS defined with parameter NORETRIEVE, which stops these migrated datasets being recalled to disk when mounts are initiated. Again, this is the required behaviour, changing this mainframe NFS parameter would see all migrated datasets referenced by the Docker NFS mount to be recalled to disk, even thought they are not being read or required. This would simply just get us over the Docket container start up, but is not idea to say the least.

Once the Docker container has started, we could then re-migrate said datasets. However this can cause space issues with our SMS pooled DASD and goes against the whole ethos of migrating datasets in the first place. It would be a very dirty ‘working around’ rather than a proper/smart resolution.

There are no further options under the mainframe NFS task to assist here as this is working as designed with regard to migrated datasets (so IBM have advised). The issue lies with the fact the Docker container cannot ‘handle’ this ‘device not available’ error, which is being thrown because ‘some’ datasets references by the mount are in a migrated state.

What can be done either within the Docker container or even with NFS/on mainframe (if there is an option we are not aware of) to ‘handle’ or resolve this issue?

Hopefully someone out there can help us and has experience of this issue and how to resolve it or get around it with a little more finesse that recalling all migrated datasets and re-migrating them once the container successfully starts.

Thanks all.