Allocate channel for RMAN is used in various scenarios, most of the time is useful when you use tape as device type or you need to use some kind of format. The way to do the allocation not changed since a long time ago, but when you run the database in container mode you can hit a bug that turns your channel unusable. I will show you the bug and how to avoid it with a simple trick.
This bug hit every version since 12 and I discovered it last year when testing some scenarios, but I was able to test and post just recently. It just occurs for CDB databases and exists just one one-off solution published for 12.2. But there is one workaround more useful and works for every version.
The most interesting part is that everything that we made until now when allocate channel will not work. You can search in all doc available for allocate channel since 9i until 19c the first thing that you made after open the run{} is allocate channel. This is the default and recommended in the docs: for 19c, 18c, and 9i.