The situation of presenting multipath storage into the containers using lxc.mount commands for /dev and /dev/mapper (so that the links are valid from /dev/mapper --> /dev/dm-*) causes some issues that are not entirely addressed in the current implementations of LXC. Therefore, it will be best to keep those commands in the container config but to have them commented out be default and only turned on if the user wants to present LUNs into the LXC container for various purposes (such as Oracle RDBMS ASM LUNs for example).
Also need to see if there is a way to use the NAME parameter in UDEV rules still, or something similar, to create actual device nodes for storage (not symlinks) in /dev/containername so that all of /dev does not have to be presented to the container and so that links are not needed (/dev AND /dev/mapper).
The situation of presenting multipath storage into the containers using lxc.mount commands for /dev and /dev/mapper (so that the links are valid from /dev/mapper --> /dev/dm-*) causes some issues that are not entirely addressed in the current implementations of LXC. Therefore, it will be best to keep those commands in the container config but to have them commented out be default and only turned on if the user wants to present LUNs into the LXC container for various purposes (such as Oracle RDBMS ASM LUNs for example).
Also need to see if there is a way to use the NAME parameter in UDEV rules still, or something similar, to create actual device nodes for storage (not symlinks) in /dev/containername so that all of /dev does not have to be presented to the container and so that links are not needed (/dev AND /dev/mapper).
This discussion here is relevant to this issue: https://github.com/lxc/lxc/issues/1552