Skip to content

Latest commit

 

History

History
378 lines (245 loc) · 19.9 KB

CHANGELOG.md

File metadata and controls

378 lines (245 loc) · 19.9 KB
Raw

Changelog

This changelog is continued from enterprise-metrics after Grafana Enterprise Metrics was added to mimir-distributed in PR #1203. All notable changes to this chart will be documented in this file.

Entries should be ordered as follows:

  • [CHANGE]
  • [FEATURE]
  • [ENHANCEMENT]
  • [BUGFIX]

Entries should include a reference to the Pull Request that introduced the change.

main / unreleased

  • [CHANGE] Enable multi-tenancy by default. This means multitenancy_enabled is now true for both Mimir and Enterprise Metrics. Nginx will inject X-Scope-OrgID=anonymous header if the header is not present, ensuring backwards compatibility. #2117
  • [CHANGE] breaking change Chart now uses custom memcached templates to remove bitnami dependency. There are changes to the Helm values, listed bellow. #2064
    • The memcached section now contains common values shared across all memcached instances.
    • New memcachedExporter section was added to configure memcached metrics exporter.
    • New memcached-chunks section was added that refers to previous memcached configuration.
    • The section memcached-queries is renamed to memcached-index-queries.
    • The value memcached-*.replicaCount is replaced with memcached-*.replicas to align with the rest of the services.
      • Renamed memcached.replicaCount to memcached-chunks.replicas.
      • Renamed memcached-queries.replicaCount to memcached-index-queries.replicas.
      • Renamed memcached-metadata.replicaCount to memcached-metadata.replicas.
      • Renamed memcached-results.replicaCount to memcached-results.replicas.
    • All memcached instances now share the same ServiceAccount that the chart uses for its services.
    • The value memcached-*.architecture was removed.
    • The value memcached-*.arguments was removed, the default arguments are now encoded in the template. Use memcached-*.extraArgs to provide additional arguments as well as the values memcached-*.allocatedMemory, memcached-*.maxItemMemory and memcached-*.port to set the memcached command line flags -m, -I and -u.
    • The remaining arguments are aligned with the rest of the chart's services, please consult the values file to check whether a parameter exists or was renamed.
  • [CHANGE] Change default value for blocks_storage.bucket_store.chunks_cache.memcached.timeout to 450ms to increase use of cached data. #2035
  • [FEATURE] Add mimir-continuous-test in smoke-test mode. Use helm test to run a smoke test of the read + write path.
  • [ENHANCEMENT] ServiceMonitor object will now have default values based on release namesapce in the namespace and namespaceSelector fields. #2123
  • [ENHANCEMENT] Set the namespace metadata field for all kubernetes objects to enable using --namespace correctly with Helm even if the specified namespace does not exist. #2123
  • [ENHANCEMENT] The new value serviceMonitor.clusterLabel controls whether to add a cluster label and with what content to ServiceMonitor metrics. #2125
  • [ENHANCEMENT] Set the flag ingester.ring.instance-availability-zone to zone-default for ingesters. This is the first step of introducing multi-zone ingesters. #2114
  • [ENHANCEMENT] Add mimir.structuredConfig for adding and modifing mimir.config values after template evaulation. It can be used to alter individual values in the configuration and it's structured YAML instead of text. #2100
  • [ENHANCEMENT] Add global.podAnnotations which can add POD annotations to PODs directly controlled by this chart (mimir services, nginx). #2099
  • [ENHANCEMENT] Introduce the value configStorageType which can be either ConfigMap or Secret. This value sets where to store the Mimir/GEM application configuration. When using the value ConfigMap, make sure that any secrets, passwords, keys are injected from the environment from a separate Secret. See also: #2031, #2017. #2089
  • [ENHANCEMENT] Add global.extraEnv and global.extraEnvFrom to values. This enables setting common environment variables and common injection of secrets to the POD environment of Mimir/GEM services and Nginx. Memcached and minio are out of scope for now. #2031
  • [ENHANCEMENT] Add extraEnvFrom capability to all Mimir services to enable injecting secrets via environment variables. #2017
  • [ENHANCEMENT] Enable -config.expand-env=true option in all Mimir services to be able to take secrets/settings from the environment and inject them into the Mimir configuration file. #2017
  • [ENHANCEMENT] Add a simple test for enterprise installation #2027
  • [ENHANCEMENT] Check for the containerSecurityContext in values file. #2112
  • [ENHANCEMENT] Add NOTES.txt to show endpoints URLs for the user at install/upgrade. #2189

2.1.0-beta.7

2.1.0-beta.6

  • [ENHANCEMENT] Disable ingester.ring.unregister-on-shutdown and distributor.extend-writes #1994
    • This will prevent resharding every series during a rolling ingester restart
    • Under some circumstances the previous values (both enabled) could cause write path degredation during rolling restarts

2.1.0-beta.5

  • [ENHANCEMENT] Add support for the results cache used by the query frontend #1993
    • This will result in additional resource usage due to the addition of one or more memcached replicas. This applies when using small.yaml, large.yaml, capped-large.yaml, capped-small.yaml, or when setting memcached-results.enabled=true

2.1.0-beta.4

  • [BUGFIX] Set up using older bitnami chart repository for memcached as old charts were deleted from the current one. #1998

2.1.0-beta.3

  • [BUGFIX] Use grpc round-robin for distributor clients in GEM gateway and self-monitoring
    • This utilizes an additional headless service for the distributor pods

2.1.0-beta.2

  • [ENHANCEMENT] Version bump only for release tests.

2.1.0-beta.1

  • [ENHANCEMENT] Version bump only for release tests.

2.0.14

  • [BUGFIX] exclude headless services from ServiceMonitors to prevent duplication of prometheus scrape targets #1308

2.0.13

  • [ENHANCEMENT] Removed rbac.create option. #1317

2.0.12

  • [ENHANCEMENT] Add memberlist named port to container spec. #1311

2.0.11

  • [ENHANCEMENT] Turn ruler and override-exporter into optional components. #1304

2.0.10

  • [ENHANCEMENT] Reorder some values for consistency. #1302
  • [BUGFIX] Add missing admin_api.env, gateway.env and overrides_exporter.env values. #1302
  • [BUGFIX] Remove <service>.extraPorts from values as it has no effect. #1302

2.0.9

  • [ENHANCEMENT] Disable gateway ingress by default. #1303
  • [BUGFIX] Fix null port at gateway ingress definition. #1303

2.0.8

  • [ENHANCEMENT] Add validation if activity_tracker.filepath is missing in mimir.config. #1290
  • [ENHANCEMENT] Add validation if server.http_listen_port or server.grpc_listen_port is set in mimir.config. #1290
  • [BUGFIX] Add missing empty array definition for extraVolumeMounts in admin_api, gateway and override-exporter. #1290
  • [BUGFIX] Fix wrong template called in nginx helper. #1290

2.0.7

  • [ENHANCEMENT] Add option to modify the port for the GEM gateway service. #1270

2.0.6

  • [ENHANCEMENT] Add option for an ingress on GEM gateway. #1266

2.0.5

  • [BUGFIX] Use new component name system for gateway ingress. This regression has been introduced with #1203. #1260

2.0.4

  • [ENHANCEMENT] Determine PodDisruptionBudget APIVersion based on running version of k8s #1229

2.0.3

  • [ENHANCEMENT] Update README.md with helm-docs version 1.8.1 instead of old 1.4.0. #1230

2.0.2

  • [ENHANCEMENT] Update Grafana Enterprise Metrics docker image tag to v2.0.1 #1241

2.0.1

  • [BUGFIX] Honor global.clusterDomain when referencing internal services, e.g. alertmanager or nginx gateway. #1227

2.0.0

  • [CHANGE] Breaking for existing users of mimir-distributed: the naming convention is changed to have shorter resource names, as in <release>-mimir-distributed-store-gateway is now just <release>-mimir-store-gateway. To have the previous names, please specify nameOverride: mimir-distributed in the values. #1203
  • [CHANGE] The chart enterprise-metrics is renamed to mimir-distributed. #1203
  • [CHANGE] Breaking Configuration for Grafana Enterprise Metrics is now in the value mimir.config as a helm template string. Please consult the Grafana Enterprise Migration Guide to learn more about how to upgrade the configuration. Except for the following parameters specified as command line parameters in the Pod templates, everything is now set in this string-typed value, giving a definitive source of configuration. Exceptions:

    The -target= must be provided individually.
    The -config.file= obviously.
    User defined arguments from .<service>.extraArgs.

  • [CHANGE] Breaking Kubernetes object labels now follow the kubernetes standard (e.g. app.kubernetes.io/component=ingester). To enable smooth upgrade and compatibility with previous Grafana Enterprise Metrics Helm chart, the value enterprise.legacyLabels should be set to true.
  • [CHANGE] Breaking Ingesters only support StatefulSet from now on as chunks storage was removed in favour of blocks storage.
  • [CHANGE] Breaking Compactor is a required component, the value compactor.enabled is removed.
  • [CHANGE] Breaking The configuration parameter server.http_listen_port and server.grpc_listen_port cannot be changed from their defaults.
  • [CHANGE] The default for ingester.ring.replication_factor is now 3 and there will be 3 ingesters started even with the default values.yaml. On the other hand, Pod anti affinity is turned off by default to allow single node deployment.
  • [FEATURE] Upgrade to Grafana Enterprise Metrics v2.0.0
  • [FEATURE] Reworked chart to enable installing Grafana Mimir open source software version without licensed features.
  • [FEATURE] Added the value nameOverride to enable migration from Cortex helm chart.
  • [FEATURE] The alertmanager can be disabled with alertmanager.enabled: false, to support the use case of external alertmanager.
  • [FEATURE] Added definitions of ServiceMonitor objects for Prometheus monitoring. Configuration is done via the serviceMonitor values. This enables partial functionality of Grafana Mimir dashboards out of the box - without alerts and recording rules pre-loaded.
  • [ENHANCEMENT] Minio bucket creation is not tied to admin-api anymore, moved to its own job templates/minio/create-bucket-job.yaml.
  • [BUGFIX] .<service>.PodDisruptionBudget was not working. Added template definition for all services. Pod disruption budget is enabled for the ingesters and store-gateways by default.
  • [BUGFIX] Fix typo in value .alertmanager.statefulset to .alertmanager.statefulSet.
  • [BUGFIX] Remove unused value .useExternalLicense.

Entries from enterprise-metrics chart

1.8.1

  • [ENHANCEMENT] Support Grafana Mimir monitoring mixin labels by setting container names to the component names. This will make it easier to select different components in cadvisor metrics. Previously, all containers used "enterprise-metrics" as the container name. Now, for example, the ingester Pod will have a container name "ingester" rather than "enterprise-metrics".

1.8.0

1.7.3

  • [BUGFIX] Alertmanager does not fail anymore to load configuration via the API. #945

1.7.2

  • [CHANGE] The Ingester statefulset now uses podManagementPolicy Parallel, upgrading requires recreating the statefulset #920

1.7.1

  • [BUGFIX] Remove chunks related default limits. #867

1.7.0

1.6.0

  • [FEATURE] Upgrade to Grafana Enterprise Metrics v1.5.1. #729
  • [CHANGE] Production values set the ingester replication factor to three to avoid data loss. The resource calculations of these values already factored in this replication factor but did not apply it in the configuration. If you have not reduced the compute resources in these values then this change should have no impact besides increased resilience to ingester failure. If you have reduced the compute resources, consider increasing them back to the recommended values before installing this version. #729

1.5.6

  • [BUGFIX] YAML exports are no longer included as part of the Helm chart. #726

1.5.5

  • [BUGFIX] Ensure all PodSpecs have configurable initContainers. #708

1.5.4

  • [BUGFIX] Adds a Service resource for the Compactor Pods and adds Compactor to the default set of gateway proxy URLs. In previous chart versions the Compactor would not show up in the GEM plugin "Ring Health" tab because the gateway did not know how to reach Compactor. #714

1.5.3

  • [BUGFIX] This change does not affect single replica deployments of the admin-api but does fix the potential for an inconsistent state when running with multiple replicas of the admin-api and experiencing parallel writes for the same objects. #675

1.5.2

  • [CHANGE] Removed all references to Consul in the yaml files since GEM will be focused on deploying with memberlist. Deleted the multi-kv-consul-primary-values.yaml and multi-kv-memberlist-primary-values.yaml files since they assume you're running Consul as your primary or second kvstore. #674

1.5.1

  • [BUGFIX] Unused ingress configuration section removed from values.yaml. #658

1.5.0

1.4.7

  • [CHANGE] Enabled enterprise authentication by default.

    Breaking: This change can cause losing access to the GEM cluster in case auth.type has not been set explicitly. This is a security related change and therefore released in a patch release.

1.4.6

  • [FEATURE] Run an instance of the GEM overrides-exporter by default. #590

1.4.5

  • [BUGFIX] Add memberlist.join configuration to the ruler. #618

1.4.4

  • [CHANGE] Removed livenessProbe configuration as it can often be more detrimental than having none. Users can still configure livenessProbes with the per App configuration hooks. #594

1.4.3

  • [ENHANCEMENT] Added values files for installations that require setting resource limits. #583

1.4.2

  • [CHANGE] The compactor data directory configuration has been corrected to /data. #562

    Note: The compactor is stateless and no data stored in the existing data directory needs to be moved in order to facilitate this upgrade. For more information, refer to the Cortex Compactor documentation.

  • [FEATURE] Upgrade to Grafana Enterprise Metrics v1.4.2 #562

1.4.1

  • [BUGFIX] Fixed DNS address of distributor client for self-monitoring. #569

1.4.0

  • [CHANGE] Use updated querier response compression configuration, changed in 1.4.0. #524
  • [CHANGE] Use updated alertmanager storage configuration, changed in 1.4.0. #524
  • [FEATURE] Upgrade to Grafana Enterprise Metrics v1.4.1. #524
  • [FEATURE] Enable GEM self-monitoring. #524

1.3.5

  • [CHANGE] The GRPC port on the query-frontend and store-gateway Kubernetes Services have been changed to match the naming of all other services. #523
  • [FEATURE] Expose GRPC port on all GEM services. #523

1.3.4

  • [BUGFIX] Removed symlinks from chart to fix Rancher repository imports. #504

1.3.3

  • [FEATURE] The GEM config now uses the {{ .Release.Name }} variable as the default value for cluster_name which removes the need to additionally override this setting during an initial install. #500

1.3.2

  • [FEATURE] Chart memcached dependencies are now at the latest release. This includes the memcached and the related exporter. #467

1.3.1

  • [BUGFIX] Use non-deprecated alertmanager flags for cluster peers. #441
  • [BUGFIX] Make store-gateway Service not headless. #441

1.3.0

1.2.0

  • [CHANGE] The chart now uses memberlist for the ring key-value store removing the need to run Consul. #340

    Warning: Existing clusters will need to follow an upgrade procedure. Warning: Existing clusters should first be upgraded to v1.1.1 and use that version for migration before upgrading to v1.2.0. To upgrade to using memberlist:

    1. Ensure you are running the v1.1.1 version of the chart.
    2. Deploy runtime multi_kv_config to use Consul as a primary and memberlist as the secondary key-value store. The values for such a change can be found in the multi-kv-consul-primary-values.yaml.
    3. Verify the configuration is in use by querying the Configuration HTTP API endpoint.
    4. Deploy runtime multi_kv_config to use memberlist as the primary and Consul as the secondary key-value store. The values for such a change can be found in multi-kv-memberlist-primary-values.yaml
    5. Verify the configuration is in use by querying the Configuration HTTP API endpoint.
    6. Deploy v1.2.0 helm chart which configures memberlist as the sole key-value store and removes the Consul resources.

1.1.1

1.1.0

  • [CHANGE] The memcached chart from the deprecated Helm stable repository has been removed and replaced with a Bitnami chart. #333

    Warning: This change will result in the cycling of your memcached Pods and will invalidate the existing cache.

  • [CHANGE] Memcached Pod resource limits have been lowered to match requests. #333
  • [FEATURE] YAML exports have been created for all chart values files. #333
  • [BUGFIX] The values for the querier/ruler/store-gateway -<prefix>.memcached.max-item-size have been corrected to match the limit configured on the memcached server. #333

1.0.0

  • [FEATURE] Initial versioned release. ##168

Entries from mimir-distributed chart

0.1.8

  • [BUGFIX] Fix nginx routing for rules and expose buildinfo. #1233

0.1.7

  • [BUGFIX] Remove misplaced config value and add affinity rules in capped-small.yaml and capped-large.yaml. #1225

0.1.6

  • [CHANGE] Breaking Compactor is a required component, the value compactor.enabled is removed. #1193
  • [FEATURE] The alertmanager can be disabled with alertmanager.enabled: false, to support the use case of external alertmanager. #1193

0.1.5

  • [BUGFIX] Fix labels for Mimir dashboards. #1190

0.1.4

  • [BUGFIX] Fix documentation link missing slash. #1177

0.1.3

  • [FEATURE] Add ServiceMonitor definitions. #1156

0.1.2

  • [BUGFIX] Fix the naming of minio configmap and secret in the parent chart. #1152

0.1.1

  • [BUGFIX] CI fixes. #1144

0.1.0

  • [FEATURE] Initial commit, Mimir only, derived from enterprise-metrics chart. #1141