Hi All - I’m creating a PR to add additional arbitrary container args to the docker run launch of action containers, and before I add additional functionality to WhiskConfig, wanted to ask about moving existing WhiskConfig vars to PureConfig.
AFAIK, if I refactor the existing WhiskConfig.invokerContainerNetwork, and replace it with, say, ContainerPool.ContainerArgsConfig.network, this will imply that invoker deployments would need to switch from using INVOKER_CONTAINER_NETWORK env var to using CONFIG_whisk_containerPool_containerArgs_network env var. I haven’t seen any precedent for doing this, but one option is to - set the old var (still) using the existing value - to set the new var, use CONFIG_whisk_containerPool_network=$INVOKER_CONTAINER_NETWORK, until a deployment is updated to have set the CONFIG_ var directly. (See [1] for details) In case there are any objections, let me know, otherwise I’ll create a PR. I think the only question is whether it is better to have this double-config var situation for some amount of time, or just force deployments to cut over to using a new environment variable name. Obviously the former is more graceful, but it will leave a trail of extra vars. This is sort of an “env var name deprecation”. Thanks Tyson [1] https://github.com/apache/incubator-openwhisk/pull/3076
