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 Docket NFS mount to be recalled to disk, even thought they are not being read or required. This would simple just get us over the Docket container start up.
Once the Docker container has started, we could re-migrate said datasets. However this can cause space issues with out 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 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. 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 within the Docker container 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.