ServiceD

Goals and Mission

We want a central access point for

• Service Management
• Service Dependency Visualization
• Service Discovery
• Service Monitoring
• Incident Management
• Feature Views

Service Management

Many services come and go. This is a central place to declare a service with all its meta information and how to handle it.

We can deploy, monitor, and revert your service in the cloud.

Internals

• service name (e.g. frontend)
• list of hard service-dependencies
• list of soft service-dependencies
• app spec: a JSON blob, currently directly matching the Marathon JSON blob

Service Dependency Visualization

Most service depend on other services. The more services get, the more complex the application stack as a whole will become.

When declaring a new service all dependencies must be referenced, too.

This way we can provide you automatically with service dependency graphs you can walk though and easily identify architectual bottlenecks and it can aid in incident management to identify dependant services that are affected by other services downtime.

Service Discovery

With many services and even more hardware resources we need a way to generically tell the application what service to find via which endpoint.

Service Monitoring

A service or feature should go into production when its monitoring is set in place. So we couple them tightly together. When you declare a service with its dependencies, you also declare health checks and responsibilities (contacts).

A service can only be deployed when it is fully declared, including health checks.

Contacts help you to get notified via SMS as soon as a service fails. Escalation contacts are notified when the service is not marked as behind repaired within a per-service customizable time frame.

Incident Management

Automatically pre-fill and issue incident reports with downtime durations, affected services and customer facing features.

Feature Views

Feature Views provide a high level abstract overview of lower level services. These Features are used by managers and customer support to know what systems are affected in a potential incident to gain more speed in communication channels.