The underlying issue regarding the dots does not seem to come from MongoDB because we're using version 3.6.3. The issue seems to be about the loss of detail when unmarshaling JSON into the Go structs. Even though the JSON tagging of these structs specifies if there's a dot or a slash (yes, slashes are also a problem), that tagging is not used when writing the structs into MongoDB. Consequently, when DCM reads from the database during /apply, it won't be aware of which fields have dots or slash in their names while constructing the yaml to be stored in etcd and consumed by rsync. This is where the loss of detail happens. Significant rework seems to be needed to fix this (perhaps the EMCO/Go long-timers can chime in and suggest a quick fix).

As such, I'll close the task now [gRPC calling is working and all resources are being created during /apply and destroyed during /terminate]. The only gap is due to the issue above so I am not adding any ResourceQuota attributes that have dots or slashes in their names (like limits.cpu). A future story will be created to specifically address this problem.

Increased scope of story:

• Pre-process all k8s resources that include a dot (.) in their key name as dots are not supported in MongoDB thus considerably breaking things and preventing resource quotas (e.g.) from actually getting created in Kubernetes.

Can now call rsync and I saw the very first k8s resource automatically create by rsync as intended by dcm. A namespace. Will close task once all basic resources can get created by rsync (namespace, role, rolebinding, quota). For now only validated in 1 cluster.