deploy/otel/col/UPGRADING.md

Upgrade guidelines

0.46.0 to 0.47.0

Update Collector Endpoints to use Pod IP Instead of 0.0.0.0

The Collector's security guidelines were updated to include containerized environments when discussing safeguards against denial of service attacks. To be in compliance with the Collector's security best practices the chart has been updated to use the Collector's pod IP in place of 0.0.0.0.

The chart will continue to allow complete configuration of the Collector via the config field in the values.yaml. If pod IP does not suite your needs you can use config to set something different.

See Security Best Practices docummentation for more details.

0.40.7 to 0.41.0

Require Kubernetes version 1.23 or later

If you enable use of a HorizontalPodAutoscaler for the collector when running in the "deployment" mode by way of .Values.autoscaling.enabled, the manifest now uses the "autoscaling/v2" API group version, which is available only as recently as Kubernetes version 1.23. As all previous versions of this API group are deprecated and removed as of Kubernetes version 1.26, we don't offer support for Kubernetes versions older than 1.23.

0.34.0 to 0.34.0

config supports templating

The chart now supports templating in .Values.config. If you are currently using any {{ }} syntax in .Values.yaml it will now be rendered. To escape existing instances of {{ }}, use {{` <original content> `}}. For example, {{ REDACTED_EMAIL }} becomes {{` {{ REDACTED_EMAIL }} `}}.

0.28.0 to 0.29.0

Reduce requested resources

Resource limits have been reduced. Upgrades/installs of chart 0.29.0 will now use fewer resources. In order to set the resources back to what they were, you will need to override the resources section in the values.yaml.

Example:

resources:
  limits:
    cpu: 1
    memory: 2Gi

0.23.1 to 0.24.0

[Remove containerLogs in favor of presets.logsCollection]()

The ability to enable logs collection from the collector has been moved from containerLogs.enabled to presets.logsCollection.enabled. If you are currently using containerLogs.enabled, you should instead use the preset:

presets:
  logsCollection:
    enabled: true

If you are using containerLogs.enabled and also enabling collection of the collector logs you can use includeCollectorLogs

presets:
  logsCollection:
    enabled: true
    includeCollectorLogs: true

You no longer need to update config.service.pipelines.logs to include the filelog receiver yourself as the preset will automatically update the logs pipeline to include the filelog receiver.

The filelog's preset configuration can modified by config.receivers, but preset configuration cannot be removed. If you need to remove any filelog receiver configuration generated by the preset you should not use the preset. Instead, configure the filelog receiver manually in config.receivers and set any other necessary fields in the values.yaml to modify k8s as needed.

See the daemonset-collector-logs example to see an example of the preset in action.

0.18.0 to 0.19.0

Remove agentCollector and standaloneCollector settings

The agentCollector and standaloneCollector config sections have been removed. Upgrades/installs of chart 0.19.0 will fail if agentCollector or standaloneCollector are in the values.yaml. See the Migrate to mode steps for instructions on how to replace agentCollector and standaloneCollector with mode.

0.13.0 to 0.14.0

Remove two-deployment mode

The ability to install both the agent and standalone collectors simultaneous with the chart has been removed. Installs/upgrades where both .Values.agentCollector.enabled and .Values.standloneCollector.enables are true will fail. agentCollector and standloneCollector have also be deprecated, but backward compatibility has been maintained.

To run both a deployment and daemonset

Install a deployment version of the collector. This is done by setting .Values.mode to deployment

mode: deployment

Next, install an daemonset version of the collector that is configured to send traffic to the previously installed deployment. This is done by setting .Values.mode to daemonset and updating .Values.config so that data is exported to the deployment.

mode: daemonset

config:
  exporters:
    otlp:
      endpoint: example-opentelemetry-collector:4317
      tls:
        insecure: true
  service:
    pipelines:
      logs:
        exporters:
         - otlp
         - logging
      metrics:
        exporters:
         - otlp
         - logging
      traces:
        exporters:
         - otlp
         - logging

See the daemonset-and-deployment example to see the rendered config.

Migrate to mode:

The agentCollector and standaloneCollector sections in values.yaml have been deprecated. Instead there is a new field, mode, that determines if the collector is being installed as a daemonset or deployment.

# Valid values are "daemonset" and "deployment".
# If set, agentCollector and standaloneCollector are ignored.
mode: <daemonset|deployment>

The following fields have also been added to the root-level to replace the depracated agentCollector and standaloneCollector settings.

containerLogs:
  enabled: false

resources:
  limits:
    cpu: 1
    memory: 2Gi

podAnnotations: {}

podLabels: {}

# Host networking requested for this pod. Use the host's network namespace.
hostNetwork: false

# only used with deployment mode
replicaCount: 1

annotations: {}

When using mode, these settings should be used instead of their counterparts in agentCollector and standaloneCollector.

Set mode to daemonset if agentCollector was being used. Move all agentCollector settings to the corresponding root-level setting. If agentCollector.configOverride was being used, merge the settings with .Values.config.

Example agentCollector values.yaml:

agentCollector:
  resources:
    limits:
      cpu: 3
      memory: 6Gi
  configOverride:
    receivers:
      hostmetrics:
        scrapers:
          cpu:
          disk:
          filesystem:
    service:
      pipelines:
        metrics:
          receivers: [otlp, prometheus, hostmetrics]

Example mode values.yaml:

mode: daemonset

resources:
  limits:
    cpu: 3
    memory: 6Gi

config:
  receivers:
    hostmetrics:
      scrapers:
        cpu:
        disk:
        filesystem:
  service:
    pipelines:
      metrics:
        receivers: [otlp, prometheus, hostmetrics]

Set mode to deployment if standaloneCollector was being used. Move all standaloneCollector settings to the corresponding root-level setting. If standaloneCollector.configOverride was being used, merge the settings with .Values.config.

Example standaloneCollector values.yaml:

standaloneCollector:
  enabled: true
  replicaCount: 2
  configOverride:
    receivers:
      podman_stats:
        endpoint: unix://run/podman/podman.sock
        timeout: 10s
        collection_interval: 10s
    service:
      pipelines:
        metrics:
          receivers: [otlp, prometheus, podman_stats]

Example mode values.yaml:

mode: deployment

replicaCount: 2

config:
  receivers:
    receivers:
      podman_stats:
        endpoint: unix://run/podman/podman.sock
        timeout: 10s
        collection_interval: 10s
  service:
    pipelines:
      metrics:
        receivers: [otlp, prometheus, podman_stats]

Default configuration in .Values.config can now be removed with null. When changing a pipeline, you must explicitly list all the components that are in the pipeline, including any default components.

Example: Disable metrics and logging pipelines and non-otlp receivers:

config:
  receivers:
    jaeger: null
    prometheus: null
    zipkin: null
  service:
    pipelines:
      traces:
        receivers:
          - otlp
      metrics: null
      logs: null