Running and deployment

This section is intended for operators running Emissary, and covers various aspects of deploying and configuring the Emissary in production.

Emissary and Kubernetes

Emissary relies on Kubernetes for reliability, availability, and scalability. This means that features such as Kubernetes readiness and liveness probes, rolling updates, and the Horizontal Pod Autoscaling should be utilized to manage Emissary.

Default configuration

The default configuration of Emissary includes default resource limits, as well as readiness and liveness probes. These values should be adjusted for your specific environment.

Running as non-root

Starting with Emissary 0.35, we support running Emissary as non-root. This is the recommended configuration, and will be the default configuration in future releases. We recommend you configure Emissary to run as non-root as follows:

  • Have Kubernetes run Emissary as non-root. This may happen by default (e.g., OpenShift) or you can set a securityContext in your Deployment as shown below in this abbreviated example:
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ambassador
spec:
  replicas: 1
  selector:
    matchLabels:
      service: ambassador
  template:
    metadata:
      labels:
        service: ambassador
    spec:
      containers:
        image: docker.io/datawire/aes:$version$
        name: ambassador
     restartPolicy: Always
     securityContext:
       runAsUser: 8888
     serviceAccountName: ambassador
  • Set the service_port element in the ambassador Module to 8080 (cleartext) or 8443 (TLS). This is the port that Emissary will use to listen to incoming traffic. Note that any port number above 1024 will work; Emissary will use 8080/8443 as its defaults in the future.

  • Make sure that incoming traffic to Emissary is configured to route to the service_port. If you’re using the default Emissary configuration, this means configuring the targetPort to point to the service_port above.

  • If you are using redirect_cleartext_from, change the value of this field to point to your cleartext port (e.g., 8080) and set service_port to be your TLS port (e.g., 8443).

Changing the configuration directory

While running, Emissary needs to use a directory within its container for generated configuration data. Normally this is /ambassador, but in some situations - especially if running as non-root - it may be necessary to change to a different directory. To do so, set the environment variable AMBASSADOR_CONFIG_BASE_DIR to the full pathname of the directory to use, as shown below in this abbreviated example:

env:
- name: AMBASSADOR_CONFIG_BASE_DIR
  value: /tmp/ambassador-config

With AMBASSADOR_CONFIG_BASE_DIR set as above, Emissary will create and use the directory /tmp/ambassador-config for its generated data. (Note that, while the directory will be created if it does not exist, attempts to turn an existing file into a directory will fail.)

Running as DaemonSet

Emissary can be deployed as a DaemonSet to have one pod per node in a Kubernetes cluster. This setup is especially helpful when you have a Kubernetes cluster running on a private cloud.

  • In an ideal example scenario, you are running containers on Kubernetes alongside with your non-containerized applications running exposed via VIP using BIG-IP or similar products. In such cases, east-west traffic is routed based on iRules to certain a set of application pools consisting of application or web servers. In this setup, alongside traditional application servers, two or more Emissary pods can also be part of the application pools. In case of failure there is at least one Emissary pod available to BIG-IP that can take care of routing traffic to the Kubernetes cluster.

  • In manifest files kind: Deployment needs to be updated to kind: DaemonSet and replicas should be removed in spec section.

Namespaces

Emissary supports multiple namespaces within Kubernetes. To make this work correctly, you need to set the AMBASSADOR_NAMESPACE environment variable in Emissary’s container. By far the easiest way to do this is using Kubernetes’ downward API (this is included in the YAML files from getambassador.io):

env:
- name: AMBASSADOR_NAMESPACE
  valueFrom:
    fieldRef:
      fieldPath: metadata.namespace

Given that AMBASSADOR_NAMESPACE is set, Emissary Mappings can operate within the same namespace, or across namespaces. Note well that Mappings will have to explicitly include the namespace with the service to cross namespaces; see the Mapping documentation for more information.

If you want Emissary to only work within a single namespace, set AMBASSADOR_SINGLE_NAMESPACE as an environment variable.

env:
- name: AMBASSADOR_NAMESPACE
  valueFrom:
    fieldRef:
      fieldPath: metadata.namespace
- name: AMBASSADOR_SINGLE_NAMESPACE
  value: "true"

With Emissary, if you set AMBASSADOR_NAMESPACE or AMBASSADOR_SINGLE_NAMESPACE, set it in the deployment container.

If you want to set a certificate for your TLScontext from another namespace, you can use the following:

env:
- name: AMBASSADOR_SINGLE_NAMESPACE
  value: "YES"
- name: AMBASSADOR_CERTS_SINGLE_NAMESPACE
  value: "YES"
- name: AMBASSADOR_NAMESPACE
  valueFrom:
    fieldRef:
      apiVersion: v1
      fieldPath: metadata.namespace

AMBASSADOR_ID

Emissary supports running multiple $productNamePlural$ in the same cluster, without restricting a given Emissary to a single namespace. This is done with the AMBASSADOR_ID setting. In the ambassador Module, set the ambassador_id, e.g.,

---
apiVersion: getambassador.io/v3alpha1
kind:  Module
metadata:
  name:  ambassador
spec:
  ambassador_id: [ "ambassador-1" ]
  config:
    ...

Then, assign each Emissary pod a unique AMBASSADOR_ID with the environment variable as part of your deployment:

env:
- name: AMBASSADOR_ID
  value: ambassador-1

With Emissary, if you set AMBASSADOR_ID, you will need to set it in the deployment container.

Emissary will then only use YAML objects that include an appropriate ambassador_id attribute. For example, if Emissary is given the ID ambassador-1 as above, only the first two YAML objects below will be used:

---
apiVersion: getambassador.io/v3alpha1
kind:  Mapping
metadata:
  name:  mapping-used
spec:
  ambassador_id: [ "ambassador-1" ]
  prefix: /demo1/
  service: demo1
---
apiVersion: getambassador.io/v3alpha1
kind:  Mapping
metadata:
  name:  mapping-used-2
spec:
  ambassador_id: [ "ambassador-1", "ambassador-2" ]
  prefix: /demo2/
  service: demo2
---
apiVersion: getambassador.io/v3alpha1
kind:  Mapping
metadata:
  name:  mapping-skipped
spec:
  prefix: /demo3/
  service: demo3
---
apiVersion: getambassador.io/v3alpha1
kind:  Mapping
metadata:
  name:  mapping-skipped-2
spec:
  ambassador_id: [ "ambassador-2" ]
  prefix: /demo4/
  service: demo4

ambassador_id is always a list, and may (as shown in mapping-used-2 above) include multiple IDs to allow a given object in the configuration for multiple Emissary instances. In this case, mapping-used-2 will be included in the configuration for ambassador-1 and also for ambassador-2.

Note well that any Emissary configuration resource can have an ambassador_id included so, for example, it is fully supported to use ambassador_id to qualify the ambassador Module, TLSContext, and AuthService objects. You will need to set ambassador_id in all resources you want to use for Emissary.

If no AMBASSADOR_ID is assigned to an Emissary, it will use the ID default. If no ambassador_id is present in a YAML object, it will also use the ID default.

AMBASSADOR_ENVOY_BASE_ID

Emissary supports running side-by-side with other envoy-based projects in a single pod. An example of this is running with an istio sidecar. This is done with the AMBASSADOR_ENVOY_BASE_ID environment variable as part of your deployment:

env:
- name: AMBASSADOR_ENVOY_BASE_ID
  value: 1

If no AMBASSADOR_ENVOY_BASE_ID is provided then it will use the ID 0. For more information on the Envoy base-id option please see the Envoy command line documentation.

AMBASSADOR_VERIFY_SSL_FALSE

By default, Emissary will verify the TLS certificates provided by the Kubernetes API. In some situations, the cluster may be deployed with self-signed certificates. In this case, set AMBASSADOR_VERIFY_SSL_FALSE to true to disable verifying the TLS certificates.

AMBASSADOR_UPDATE_MAPPING_STATUS

If AMBASSADOR_UPDATE_MAPPING_STATUS is set to the string true, Emissary will update the status of every Mapping CRD that it accepts for its configuration. This has no effect on the proper functioning of Emissary itself, and can be a performance burden on installations with many Mappings. It has no effect for Mappings stored as annotations.

The default is false. We recommend leaving AMBASSADOR_UPDATE_MAPPING_STATUS turned off unless required for external systems.

AMBASSADOR_LEGACY_MODE

Setting AMBASSADOR_LEGACY_MODE to true will result in Emissary disabling certain features and reverting to older codepaths which may be better preserve certain older behaviors. Legacy mode currently has the following effects:

  • Emissary will switch back to the Emissary 1.6 input-resource validator (which can significantly increase configuration latency for Emissary installations with many resources).
  • Emissary will use the shell boot sequence that was the default up through 1.9.1, rather than the Golang boot sequence that became the default in 1.10.0.
  • AMBASSADOR_FAST_RECONFIGURE (see below) is not supported in legacy mode.

AMBASSADOR_FAST_RECONFIGURE

Setting AMBASSADOR_FAST_RECONFIGURE to “true” enables incremental reconfiguration. When enabled, Emissary will track deltas from one configuration to the next and recalculate only what is necessary to follow the change. When disabled (the default), Emissary will recompute the entire configuration at every change.

AMBASSADOR_FAST_RECONFIGURE is not supported when AMBASSADOR_LEGACY_MODE is active.

AMBASSADOR_ENVOY_API_VERSION

By default, Emissary will configure Envoy using the Envoy V3 API. Setting AMBASSADOR_ENVOY_API_VERSION to V2 tells Ambassador to use the Envoy V2 API instead. Since the Envoy V2 API is deprecated in the upstream Envoy project, we strongly recommend using V3 and, as always, providing feedback and bug reports in the Emissary GitHub repo!

Support for the Envoy V2 API and the AMBASSADOR_ENVOY_API_VERSION environment variable will be removed in Emissary 2.2.0.

Configuration from the filesystem

If desired, Emissary can be configured from YAML files in the directory $AMBASSADOR_CONFIG_BASE_DIR/ambassador-config (by default, /ambassador/ambassador-config, which is empty in the images built by Datawire). You could volume mount an external configuration directory here, for example, or use a custom Dockerfile to build configuration directly into a Docker image.

Note well that while Emissary will read its initial configuration from this directory, configuration loaded from Kubernetes annotations will replace this initial configuration. If this is not what you want, you will need to set the environment variable AMBASSADOR_NO_KUBEWATCH so that Emissary will not try to update its configuration from Kubernetes resources.

Also note that the YAML files in the configuration directory must contain the Emissary resources, not Kubernetes resources with annotations.

Log levels and debugging

The Emissary and Ambassador Edge Stack support more verbose debugging levels. If using Emissary, the diagnostics service has a button to enable debug logging. Be aware that if you’re running Emissary on multiple pods, the debug log levels are not enabled for all pods – they are configured on a per-pod basis.

If using Ambassador Edge Stack, you can adjust the log level by setting the AES_LOG_LEVEL environment variable; from least verbose to most verbose, the valid values are error, warn/warning, info, debug, and trace; the default is info.

Log format

By default, Emissary writes its own logs in a plain text format. To produce logs as JSON instead, set the AMBASSADOR_JSON_LOGGING environment variable:

env:
- name: AMBASSADOR_JSON_LOGGING
  value: "true"

Port assignments

Emissary uses some TCP ports in the range 8000-8499 internally, as well as port 8877. Third-party software integrating with Emissary should not use ports in this range on the Emissary pod.

Emissary update checks (Scout)

Emissary integrates Scout, a service that periodically checks with Datawire servers to advise of available updates. Scout also sends anonymized usage data and the Emissary version. This information is important to us as we prioritize test coverage, bug fixes, and feature development. Note that Emissary will run regardless of the status of Scout (i.e., our uptime has zero impact on your uptime.)

We do not recommend you disable Scout, since we use this mechanism to notify users of new releases (including critical fixes and security issues). This check can be disabled by setting the environment variable SCOUT_DISABLE to 1 in your Emissary deployment.

Each Emissary installation generates a unique cluster ID based on the UID of its Kubernetes namespace and its Emissary ID: the resulting cluster ID is a UUID which cannot be used to reveal the namespace name nor Emissary ID itself. Emissary needs RBAC permission to get namespaces for this purpose, as shown in the default YAML files provided by Datawire; if not granted this permission it will generate a UUID based only on the Emissary ID. To disable cluster ID generation entirely, set the environment variable AMBASSADOR_CLUSTER_ID to a UUID that will be used for the cluster ID.

Unless disabled, Emissary will also report the following anonymized information back to Datawire:

Attribute Type Description
cluster_count int total count of clusters in use
cluster_grpc_count int count of clusters using GRPC upstream
cluster_http_count int count of clusters using HTTP or HTTPS upstream
cluster_routing_envoy_rh_count int count of clusters routing using Envoy ring_hash
cluster_routing_envoy_maglev_count int count of clusters routing using Envoy maglev
cluster_routing_envoy_lr_count int count of clusters routing using Envoy least_request
cluster_routing_envoy_rr_count int count of clusters routing using Envoy round_robin
cluster_routing_kube_count int count of clusters routing using Kubernetes
cluster_tls_count int count of clusters originating TLS
custom_ambassador_id bool has the ambassador_id been changed from ‘default’?
custom_listener_port bool has the listener port been changed from 80/443?
diagnostics bool is the diagnostics service enabled?
endpoint_grpc_count int count of endpoints to which Emissary will originate GRPC
endpoint_http_count int count of endpoints to which Emissary will originate HTTP or HTTPS
endpoint_routing bool is endpoint routing enabled?
endpoint_routing_envoy_rh_count int count of endpoints being routed using Envoy ring_hash
endpoint_routing_envoy_maglev_count int count of endpoints being routed using Envoy maglev
endpoint_routing_envoy_lr_count int count of endpoints being routed using Envoy least_request
endpoint_routing_envoy_rr_count int count of endpoints being routed using Envoy round_robin
endpoint_routing_kube_count int count of endpoints being routed using Kubernetes
endpoint_tls_count int count of endpoints to which Emissary will originate TLS
extauth bool is extauth enabled?
extauth_allow_body bool will Emissary send the body to extauth?
extauth_host_count int count of extauth hosts in use
extauth_proto str extauth protocol in use (‘http’, ‘grpc’, or null if not active)
group_canary_count int count of Mapping groups that include more than one Mapping
group_count int total count of Mapping groups in use (length of the route table)
group_header_match_count int count of groups using header matching (including host and method)
group_host_redirect_count int count of groups using host_redirect
group_host_rewrite_count int count of groups using host_rewrite
group_http_count int count of HTTP Mapping groups
group_precedence_count int count of groups that explicitly set the precedence of the group
group_regex_header_count int count of groups using regex header matching
group_regex_prefix_count int count of groups using regex prefix matching
group_resolver_consul int count of groups using the Consul resolver
group_resolver_kube_endpoint int count of groups using the Kubernetes endpoint resolver
group_resolver_kube_service int count of groups using the Kubernetes service resolver
group_shadow_count int count of groups using shadows
group_shadow_weighted_count int count of groups using shadows but not shadowing all traffic
group_tcp_count int count of TCP Mapping groups
host_count int count of Host resources in use
k8s_ingress_class_count int count of IngressClass resources in use
k8s_ingress_count int count of Ingress resources in use
listener_count int count of active listeners (1 unless redirect_cleartext_from or TCP Mappings are in use)
liveness_probe bool are liveness probes enabled?
managed_by string tool that manages the Emissary deployment, if any (e.g. helm, edgectl, etc.)
mapping_count int count of Mapping resources in use
ratelimit bool is rate limiting in use?
ratelimit_custom_domain bool has the rate limiting domain been changed from ‘ambassador’?
ratelimit_data_plane_proto bool is rate limiting using the data plane proto?
readiness_probe bool are readiness probes enabled?
request_4xx_count int lower bound for how many requests have gotten a 4xx response
request_5xx_count int lower bound for how many requests have gotten a 5xx response
request_bad_count int lower bound for how many requests have failed (either 4xx or 5xx)
request_elapsed float seconds over which the request_ counts are valid
request_hr_elapsed string human-readable version of request_elapsed (e.g. “3 hours 35 minutes 20 seconds”
request_ok_count int lower bound for how many requests have succeeded (not a 4xx or 5xx)
request_total_count int lower bound for how many requests were handled in total
statsd bool is StatsD enabled?
server_name bool is the server_name response header overridden?
service_resource_total int total count of service resources loaded from all discovery sources
tls_origination_count int count of TLS origination contexts
tls_termination_count int count of TLS termination contexts
tls_using_contexts bool are new TLSContext resources in use? ?
tls_using_module bool is the old TLS module in use
tracing bool is tracing in use?
tracing_driver str tracing driver in use (‘zipkin’, ’lightstep’, ‘datadog’, or null if not active)
use_proxy_proto bool is the PROXY protocol in use?
use_remote_address bool is Emissary honoring remote addresses?
x_forwarded_proto_redirect bool is Emissary redirecting based on X-Forwarded-Proto?
xff_num_trusted_hops int what is the count of trusted hops for X-Forwarded-For?

The request_* counts are always incremental: they contain only information about the last request_elapsed seconds. Additionally, they only provide a lower bound – notably, if an Emissary pod crashes or exits, no effort is made to ship out a final update, so it’s very easy for counts to never be reported.

To completely disable feature reporting, set the environment variable AMBASSADOR_DISABLE_FEATURES to any non-empty value.


Last modified September 9, 2024: Update all 1.14 metadata to fix navigation (c0afada)