Brian Grant

Kubernetes Borg/Omega history topic 2: Borg had Machine key/value attributes that could be used in scheduling constraints. Borgmon had target labels to convey application topology, environment, and locale. But Jobs themselves didn't originally have k/v labels.

So Borg users would embed attribute values in Job names, separated by dots and dashes, up to 180 characters long, and then parse them out in other systems and tools using complex regular expressions.

It was clear that load balancers, monitoring systems, and release, rollout, and configuration tools needed common support for identifying attributes that could flow through the entire lifecycle of an application. However, single-value tags, as in gmail and GCE, were missing types

Thus, in mid 2013, we proposed key/value labels, for both Borg and Google Cloud, to facilitate higher-level application management. Of course, it was much easier to incorporate them into a greenfield project like Kubernetes, which had labels from the beginning.

Label selector semantics were designed originally for the monitoring system. Monitoring and load-balancing systems wanted to ensure that non-overlapping queries could be constructed. Without disjunction, a common key with different values ensures two selectors don't overlap