Policies and Telemetry

Istio provides a flexible model to enforce authorization policies and collect telemetry for the services in a mesh.

Infrastructure backends are designed to provide support functionality used to build services. They include such things as access control systems, telemetry capturing systems, quota enforcement systems, billing systems, and so forth. Services traditionally directly integrate with these backend systems, creating a hard coupling and baking-in specific semantics and usage options.

Istio provides a uniform abstraction that makes it possible for Istio to interface with an open-ended set of infrastructure backends. This is done in such a way to provide rich and deep controls to the operator, while imposing no burden on service developers. Istio is designed to change the boundaries between layers in order to reduce systemic complexity, eliminate policy logic from service code and give control to operators.

Mixer is the Istio component responsible for providing policy controls and telemetry collection:

Mixer Topology
Mixer Topology

The Envoy sidecar logically calls Mixer before each request to perform precondition checks, and after each request to report telemetry. The sidecar has local caching such that a large percentage of precondition checks can be performed from cache. Additionally, the sidecar buffers outgoing telemetry such that it only calls Mixer infrequently.

At a high level, Mixer provides:

  • Backend Abstraction. Mixer insulates the rest of Istio from the implementation details of individual infrastructure backends.

  • Intermediation. Mixer allows operators to have fine-grained control over all interactions between the mesh and infrastructure backends.

Beyond these purely functional aspects, Mixer also has reliability and scalability benefits as outlined below.

Policy enforcement and telemetry collection are entirely driven from configuration. Policy check is disabled by default, avoiding the need to go through the Mixer policy component. Refer to Installation Options for more information.

Adapters

Mixer is a highly modular and extensible component. One of its key functions is to abstract away the details of different policy and telemetry backend systems, allowing the rest of Istio to be agnostic of those backends.

Mixer’s flexibility in dealing with different infrastructure backends comes from its general-purpose plug-in model. Individual plug-ins are known as adapters and they allow Mixer to interface to different infrastructure backends that deliver core functionality, such as logging, monitoring, quotas, ACL checking, and more. The exact set of adapters used at runtime is determined through configuration and can easily be extended to target new or custom infrastructure backends.

Showing Mixer with adapters.
Mixer and its Adapters

Learn more about the set of supported adapters.

Reliability and latency

Mixer is a highly available component whose design helps increase overall availability and reduce average latency of services in the mesh. Key aspects of its design deliver these benefits:

  • Statelessness. Mixer is stateless in that it doesn’t manage any persistent storage of its own.

  • Hardening. Mixer proper is designed to be a highly reliable component. The design intent is to achieve > 99.999% uptime for any individual Mixer instance.

  • Caching and Buffering. Mixer is designed to accumulate a large amount of transient ephemeral state.

The sidecar proxies that sit next to each service instance in the mesh must necessarily be frugal in terms of memory consumption, which constrains the possible amount of local caching and buffering. Mixer, however, lives independently and can use considerably larger caches and output buffers. Mixer thus acts as a highly-scaled and highly-available second-level cache for the sidecars.

Mixer Topology
Mixer Topology

Since Mixer’s expected availability is considerably higher than most infrastructure backends (those often have availability of perhaps 99.9%). Mixer’s local caches and buffers not only contribute to reduce latency, they also help mask infrastructure backend failures by being able to continue operating even when a backend has become unresponsive.

Finally, Mixer’s caching and buffering helps reduce the frequency of calls to backends, and can sometimes reduce the amount of data sent to backends (through local aggregation). Both of these can reduce operational expense in certain cases.

Attributes

Attributes are an essential concept to Istio’s policy and telemetry functionality. An attribute is a small bit of data that describes a single property of a specific service request or the environment for the request. For example, an attribute can specify the size of a specific request, the response code for an operation, the IP address where a request came from, etc.

Each attribute has a name and a type. The type defines the kind of data that the attribute holds. For example, an attribute can have a STRING type which means it has a textual value, or it can have an INT64 type indicating it has a 64 bit integer value.

Here are some example attributes with their associated values:

request.path: xyz/abc
request.size: 234
request.time: 12:34:56.789 04/17/2017
source.ip: [192 168 0 1]
destination.service.name: example

Mixer is in essence an attribute processing machine. The Envoy sidecar invokes Mixer for every request, giving Mixer a set of attributes that describe the request and the environment around the request. Based on its configuration and the specific set of attributes it was given, Mixer generates calls to a variety of infrastructure backends.

Attribute Machine
Attribute Machine

Attribute vocabulary

A given Istio deployment has a fixed vocabulary of attributes that it understands. The specific vocabulary is determined by the set of attribute producers being used in the deployment. The primary attribute producer in Istio is Envoy, although specialized Mixer adapters can also generate attributes.

Learn more about the common baseline set of attributes available in most Istio deployments.

Attribute expressions

Attribute expressions are used when configuring instances. Here’s an example use of expressions:

destination_service: destination.service.host
response_code: response.code
destination_version: destination.labels["version"] | "unknown"

The sequences on the right-hand side of the colons are the simplest forms of attribute expressions. The first two only consist of attribute names. The response_code label is assigned the value from the response.code attribute.

Here’s an example of a conditional expression:

destination_version: destination.labels["version"] | "unknown"

With the above, the destination_version label is assigned the value of destination.labels["version"]. However if that attribute is not present, the literal "unknown" is used.

Refer to the attribute expression page for details.

Configuration model

Istio’s policy and telemetry features are configured through a common model designed to put operators in control of every aspect of authorization policy and telemetry collection. Specific focus was given to keeping the model simple, while being powerful enough to control Istio’s many features at scale.

Controlling the policy and telemetry features involves configuring three types of resources:

  • Configuring a set of handlers, which determine the set of adapters that are being used and how they operate. Providing a statsd adapter with the IP address for a Statsd backend is an example of handler configuration.

  • Configuring a set of instances, which describe how to map request attributes into adapter inputs. Instances represent a chunk of data that one or more adapters will operate on. For example, an operator may decide to generate requestcount metric instances from attributes such as destination.service.host and response.code.

  • Configuring a set of rules, which describe when a particular adapter is called and which instances it is given. Rules consist of a match expression and actions. The match expression controls when to invoke an adapter, while the actions determine the set of instances to give the adapter. For example, a rule might send generated requestcount metric instances to a statsd adapter.

Configuration is based on adapters and templates:

  • Adapters encapsulate the logic necessary to interface Mixer with a specific infrastructure backend.

  • Templates define the schema for specifying request mapping from attributes to adapter inputs. A given adapter may support any number of templates.

Handlers

Adapters encapsulate the logic necessary to interface Mixer with specific external infrastructure backends such as Prometheus or Stackdriver. A handler is a resource responsible for holding the configuration state needed by an adapter. For example, a logging adapter may require the IP address and port of the log collection backend.

Here is an example showing how to create a handler for an adapter. The listchecker adapter checks an input value against a list. If the adapter is configured for a whitelist, it returns success if the input value is found in the list.

apiVersion: config.istio.io/v1alpha2
kind: handler
metadata:
  name: staticversion
  namespace: istio-system
spec:
  compiledAdapter: listchecker
  params:
    providerUrl: http://white_list_registry/
    blacklist: false

The schema of the data in the params stanza depends on the specific adapter being configured.

Some adapters implement functionality that goes beyond connecting Mixer to a backend. For example, the prometheus adapter consumes metrics and aggregates them as distributions or counters in a configurable way.

apiVersion: config.istio.io/v1alpha2
kind: handler
metadata:
  name: promhandler
  namespace: istio-system
spec:
  compiledAdapter: prometheus
  params:
    metrics:
    - name: request_count
      instance_name: requestcount.instance.istio-system
      kind: COUNTER
      label_names:
      - destination_service
      - destination_version
      - response_code
    - name: request_duration
      instance_name: requestduration.instance.istio-system
      kind: DISTRIBUTION
      label_names:
      - destination_service
      - destination_version
      - response_code
      buckets:
        explicit_buckets:
          bounds: [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10]

Each adapter defines its own particular format of configuration data. Learn more about the full set of adapters and their specific configuration formats.

Instances

Instance configuration specifies the request mapping from attributes to adapter inputs. The following is an example of a metric instance configuration that produces the requestduration metric.

apiVersion: config.istio.io/v1alpha2
kind: instance
metadata:
  name: requestduration
  namespace: istio-system
spec:
  compiledTemplate: metric
  params:
    value: response.duration | "0ms"
    dimensions:
      destination_service: destination.service.host | "unknown"
      destination_version: destination.labels["version"] | "unknown"
      response_code: response.code | 200
    monitored_resource_type: '"UNSPECIFIED"'

Note that all the dimensions expected in the handler configuration are specified in the mapping. Templates define the specific required content of individual instances. Learn more about the set of templates and their specific configuration formats.

Rules

Rules specify when a particular handler is invoked with a specific instance. Consider an example where you want to deliver the requestduration metric to the prometheus handler if the destination service is service1 and the x-user request header has a specific value.

apiVersion: config.istio.io/v1alpha2
kind: rule
metadata:
  name: promhttp
  namespace: istio-system
spec:
  match: destination.service.host == "service1.ns.svc.cluster.local" && request.headers["x-user"] == "user1"
  actions:
  - handler: promhandler
    instances: [ requestduration ]

A rule contains a match predicate expression and a list of actions to perform if the predicate is true. An action specifies the list of instances to be delivered to a handler. A rule must use the fully qualified names of handlers and instances. If the rule, handlers, and instances are all in the same namespace, the namespace suffix can be elided from the fully qualified name as seen in promhandler.