1. specified cluster
2. data center as cluster
3. default cluster

1. Adjust portal jvm memory size
2. Retry if client get empty response from meta server
3. Fix potential npe issues

1. Add a release key generator to generate human-readable release keys
2. Add retry policy for client long polling
3. Move eureka url config to database

1. Refactor apollo client code according to  @qmwu2000's comments
2. Add config service integration tests
3. Add some cat events for config service
4. Fix several bugs