Description
Due to memory quotas residing in eventually consistent ns_config and being changed and checked outside of the context of the ns_orchestrator it is possible to create a node where sum of service quotas exceeds the total memory. This can be done if the quotas are changed in parallel with node joining, so node validates it's services configuration based on old quotas, and set quotas validation is performed on a cluster still without the node being joined.
The possible solution:
1. Move quotas to chronicle
2. Always check quotas in the same transaction in which node services are set.