Configuration validation fails randomly in ORY Hydra

[aeneasr]

Describe the bug

When running ORY Hydra with tracing, the service usually starts but sometimes terminates with the following error:

hydra_1          | time=2020-10-14T10:21:54Z level=error msg=The provided configuration is invalid and could not be loaded. Check the output below to understand why. audience=application config_file= service_name= service_version=
hydra_1          | 
hydra_1          | tracing.providers.jaeger.sampling: map[server_url:http://jaeger:5778/sampling type:const value:1]
hydra_1          |                                    ^-- allOf failed
hydra_1          | 
hydra_1          | tracing.providers.jaeger.sampling: map[server_url:http://jaeger:5778/sampling type:const value:1]
hydra_1          |                                    ^-- valid against schemas at indexes 0 and 1

which is non-sensical as the provided values are valid against one schema only. I have also seen other values such as <nil> for the map or values which are missing one value, e.g. map[type:const value:1].

To Reproduce

Steps to reproduce the behavior:

1. In ORY Hydra, run docker-compose -f quickstart.yml -f quickstart-postgres.yml -f quickstart-tracing.yml up
2. In another terminal, run docker restart hydra_hydra_1 or - to get the behavior consistently: for i in 1 .. 100; do docker restart hydra_hydra_1; done

Expected behavior

Obvious ;)

Environment

• ory/x 0.0.153

[aeneasr]

This is resolved by using koanf