- [BUGFIX] Fix alertmanager >=0.15.0 cluster gossip communication (#1193)
From this release onwards only Kubernetes versions v1.8 and higher are supported. If you have an older version of Kubernetes and the Prometheus Operator running, we recommend upgrading Kubernetes first and then the Prometheus Operator.
While multiple validation issues have been fixed, it will remain a beta feature in this release. If you want to update validations, you need to either apply the CustomResourceDefinitions located in example/prometheus-operator-crd
or delete all CRDs and restart the Prometheus Operator.
Some changes cause Prometheus and Alertmanager clusters to be redeployed. If you do not have persistent storage backing your data, this means you will loose the amount of data equal to your retention time.
- [CHANGE] Use canonical
/prometheus
and/alertmanager
as data dirs in containers. - [FEATURE] Allow configuring Prometheus and Alertmanager servers to listen on loopback interface, allowing proxies to be the ingress point of those Pods.
- [FEATURE] Allow configuring additional containers in Prometheus and Alertmanager Pods.
- [FEATURE] Add ability to whitelist Kubernetes labels to become Prometheus labels.
- [FEATURE] Allow specifying additonal secrets for Alertmanager Pods to mount.
- [FEATURE] Allow specifying
bearer_token_file
for Alertmanger configurations of Prometheus objects in order to authenticate with Alertmanager. - [FEATURE] Allow specifying TLS configuration for Alertmanger configurations of Prometheus objects.
- [FEATURE] Add metrics for reconciliation errors:
prometheus_operator_alertmanager_reconcile_errors_total
andprometheus_operator_prometheus_reconcile_errors_total
. - [FEATURE] Support
read_recent
andrequired_matchers
fields for remote read configurations. - [FEATURE] Allow disabling any defaults of
SecurityContext
fields of Pods. - [BUGFIX] Handle Alertmanager >=v0.15.0 breaking changes correctly.
- [BUGFIX] Fix invalid validations for metric relabeling fields.
- [BUGFIX] Fix validations for
AlertingSpec
. - [BUGFIX] Fix validations for deprecated storage fields.
- [BUGFIX] Fix remote read and write basic auth support.
- [BUGFIX] Fix properly propagate errors of Prometheus config reloader.
This release adds validations as a beta feature. It will only be installed on new clusters, existing CRD definitions will not be updated, this will be done in a future release. Please try out this feature and give us feedback!
- [CHANGE] Default Prometheus version v2.2.0-rc.0.
- [CHANGE] Default Alertmanager version v0.14.0.
- [FEATURE] Generate and add CRD validations.
- [FEATURE] Add ability to set
serviceAccountName
for Alertmanager Pods. - [FEATURE] Add ability to specify custom
securityContext
for Alertmanager Pods. - [ENHANCEMENT] Default to non-root security context for Alertmanager Pods.
- [CHANGE] Default to Alertmanager v0.13.0.
- [BUGFIX] Alertmanager flags must be double dashed starting v0.13.0.
- [FEATURE] Add support for specifying remote storage configurations.
- [FEATURE] Add ability to specify log level.
- [FEATURE] Add support for dropping metrics at scrape time.
- [ENHANCEMENT] Ensure that resource limit can't make Pods unschedulable.
- [ENHANCEMENT] Allow configuring emtpyDir volumess
- [BUGFIX] Use
--storage.tsdb.no-lockfile
for Prometheus 2.0. - [BUGFIX] Fix Alertmanager default storage.path.
- [CHANGE] Default Prometheus version v2.0.0
- [BUGFIX] Generate ExternalLabels deterministically
- [BUGFIX] Fix incorrect mount path of Alertmanager data volume
- [EXPERIMENTAL] Add ability to specify CRD Kind name
- [BUGFIX] Ignore illegal change of PodManagementPolicy to StatefulSet.
- [CHANGE] Default Prometheus version v2.0.0-rc.1.
- [CHANGE] Default Alertmanager version v0.9.1.
- [BUGFIX] Set StatefulSet replicas to 0 if 0 is specified in Alertmanager/Prometheus object.
- [BUGFIX] Glob for all files in a ConfigMap as rule files.
- [FEATURE] Add ability to run Prometheus Operator for a single namespace.
- [FEATURE] Add ability to specify CRD api group.
- [FEATURE] Use readiness and health endpoints of Prometheus 1.8+.
- [ENHANCEMENT] Add OwnerReferences to managed objects.
- [ENHANCEMENT] Use parallel pod creation strategy for Prometheus StatefulSets.
After a long period of not having broken any functionality in the Prometheus Operator, we have decided to promote the status of this project to beta.
Compatibility guarantees and migration strategies continue to be the same as for the v0.12.0
release.
- [CHANGE] Remove analytics collection.
- [BUGFIX] Fix memory leak in kubelet endpoints sync.
- [FEATURE] Allow setting global default
scrape_interval
. - [FEATURE] Allow setting Pod objectmeta to Prometheus and Alertmanger objects.
- [FEATURE] Allow setting tolerations and affinity for Prometheus and Alertmanager objects.
Starting with this release only Kubernetes v1.7.x
and up is supported as CustomResourceDefinitions are a requirement for the Prometheus Operator and are only available from those versions and up.
Additionally all objects have been promoted from v1alpha1
to v1
. On start up of this version of the Prometheus Operator the previously used ThirdPartyResource
s and the associated v1alpha1
objects will be automatically migrated to their v1
equivalent CustomResourceDefinition
.
- [CHANGE] All manifests created and used by the Prometheus Operator have been promoted from
v1alpha1
tov1
. - [CHANGE] Use Kubernetes
CustomResourceDefinition
s instead ofThirdPartyResource
s. - [FEATURE] Add ability to set scrape timeout to
ServiceMonitor
. - [ENHANCEMENT] Use
StatefulSet
rolling deployments. - [ENHANCEMENT] Properly set
SecurityContext
for Prometheus 2.0 deployments. - [ENHANCEMENT] Enable web lifecycle APIs for Prometheus 2.0 deployments.
- [BUGFIX] Fix memory leak in kubelet endpoints sync.
- [ENHANCEMENT] Add profiling endpoints.
- [BUGFIX] Adapt Alertmanager storage usage to not use deprecated storage definition.
Warning: This release deprecates the previously used storage definition in favor of upstream PersistentVolumeClaim templates. While this should not have an immediate effect on a running cluster, Prometheus object definitions that have storage configured need to be adapted. The previously existing fields are still there, but have no effect anymore.
- [FEATURE] Add Prometheus 2.0 alpha3 support.
- [FEATURE] Use PVC templates instead of custom storage definition.
- [FEATURE] Add cAdvisor port to kubelet sync.
- [FEATURE] Allow default base images to be configurable.
- [FEATURE] Configure Prometheus to only use necessary namespaces.
- [ENHANCEMENT] Improve rollout detection for Alertmanager clusters.
- [BUGFIX] Fix targetPort relabeling.
- [BUGFIX] Use computed route prefix instead of directly from manifest.
Attention: if the basic auth feature was previously used, the key
and name
fields need to be switched. This was not intentional, and the bug is not fixed,
but causes this change.
- [CHANGE] Prometheus default version v1.7.1.
- [CHANGE] Alertmanager default version v0.7.1.
- [BUGFIX] Fix basic auth secret key selector
key
andname
switched. - [BUGFIX] Fix route prefix flag not always being set for Prometheus.
- [BUGFIX] Fix nil panic on replica metrics.
- [FEATURE] Add ability to specify Alertmanager path prefix for Prometheus.
- [CHANGE] Prometheus route prefix defaults to root.
- [CHANGE] Default to Prometheus v1.7.0.
- [CHANGE] Default to Alertmanager v0.7.0.
- [FEATURE] Add route prefix support to Alertmanager resource.
- [FEATURE] Add metrics on expected replicas.
- [FEATURE] Support for runing Alertmanager v0.7.0.
- [BUGFIX] Fix sensitive rollout triggering.
- [FEATURE] Add experimental Prometheus 2.0 support.
- [FEATURE] Add support for setting Prometheus external labels.
- [BUGFIX] Fix non-deterministic config generation.
- [CHANGE] The
kubelet-object
flag has been renamed tokubelet-service
. - [CHANGE] Remove automatic relabelling of Pod and Service labels onto targets.
- [CHANGE] Remove "non-namespaced" alpha annotation in favor of
honor_labels
. - [FEATURE] Add ability make use of the Prometheus
honor_labels
configuration option. - [FEATURE] Add ability to specify image pull secrets for Prometheus and Alertmanager pods.
- [FEATURE] Add basic auth configuration option through ServiceMonitor.
- [ENHANCEMENT] Add liveness and readiness probes to Prometheus and Alertmanger pods.
- [ENHANCEMENT] Add default resource requests for Alertmanager pods.
- [ENHANCEMENT] Fallback to ExternalIPs when InternalIPs are not available in kubelet sync.
- [ENHANCEMENT] Improved change detection to trigger Prometheus rollout.
- [ENHANCEMENT] Do not delete unmanaged Prometheus configuration Secret.
- [ENHANCEMENT] Use new Prometheus 1.6 storage flags and make it default.
- [ENHANCEMENT] Include kubelet insecure port in kubelet Enpdoints object.
- [FEATURE] Add ability to mount custom secrets into Prometheus Pods. Note that secrets cannot be modified after creation, if the list if modified after creation it will not effect the Prometheus Pods.
- [FEATURE] Attach pod and service name as labels to Pod targets.
This release introduces breaking changes to the generated StatefulSet's PodTemplate, which cannot be modified for StatefulSets. The Prometheus and Alertmanager objects have to be deleted and recreated for the StatefulSets to be created properly.
- [CHANGE] Use Secrets instead of ConfigMaps for configurations.
- [FEATURE] Allow ConfigMaps containing rules to be selected via label selector.
- [FEATURE]
nodeSelector
added to the Alertmanager kind. - [ENHANCEMENT] Use Prometheus v2 chunk encoding by default.
- [BUGFIX] Fix Alertmanager cluster mesh initialization.
- [FEATURE] Allow not tagging targets with the
namespace
label. - [FEATURE] Allow specifying
ServiceAccountName
to be used by Prometheus pods. - [ENHANCEMENT] Label governing services to uniquely identify them.
- [ENHANCEMENT] Reconcile Serive and Endpoints objects.
- [ENHANCEMENT] General stability improvements.
- [BUGFIX] Hostname cannot be fqdn when syncing kubelets into Endpoints object.
- [BUGFIX] Use correct governing
Service
for PrometheusStatefulSet
.
- [FEATURE] Allow synchronizing kubelets into an
Endpoints
object. - [FEATURE] Allow specifying custom configmap-reload image
- [CHANGE] Split endpoint and job in separate labels instead of a single concatenated one.
- [BUGFIX] Properly exit on errors communicating with the apiserver.
This release introduces breaking changes to the underlying naming schemes. It is recommended to destroy existing Prometheus and Alertmanager objects and recreate them with new namings.
With this release support for v1.4.x
clusters is dropped. Changes will not be
backported to the 0.1.x
release series anymore.
- [CHANGE] Prefixed StatefulSet namings based on managing resource
- [FEATURE] Pass labels and annotations through to StatefulSets
- [FEATURE] Add tls config to use for Prometheus target scraping
- [FEATURE] Add configurable
routePrefix
for Prometheus - [FEATURE] Add node selector to Prometheus TPR
- [ENHANCEMENT] Stability improvements
- [BUGFIX] Fix config reloading when using external url.
The 0.1.x
releases are backport releases with Kubernetes v1.4.x
compatibility.
- [BUGFIX] Fix config reloading when using external url.
- [FEATURE] Add ability to set the external url the Prometheus/Alertmanager instances will be available under.
The 0.1.x
releases are backport releases with Kubernetes v1.4.x
compatibility.
- [FEATURE] Add ability to set the external url the Prometheus/Alertmanager instances will be available under.
- [BUGFIX] Fix
subPath
behavior when not using storage provisioning
This release requires a Kubernetes cluster >=1.5.0. See the readme for instructions on how to upgrade if you are currently running on a lower version with the operator.
- [CHANGE] Use StatefulSet instead of PetSet
- [BUGFIX] Fix Prometheus config generation for labels containing "-"