All notable changes to the absent metrics operator will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- New
prom-rule-name
flag which can be used to provide a template for AbsencePrometheusRule name generation and consequently absence alert rules aggregation. - Improved tests by adding dedicated unit tests for alert rule parsing and name generation edge-cases.
- Clean up of absence alert rules when a rule group is deleted.
- Heuristic determination of
tier
,service
, andsupport_group
labels. These labels will now be copied over as is from the original alert rule to its corresponding absence alert rule. tier
,service
,ccloud/service
, andccloud/support-group
labels at the object level (PrometheusRule).
- Updated all dependencies to their latest version.
- Add automaxprocs
- Updated all dependencies and golang to their latest version.
- Renamed
doc
folder todocs
.
- Update Golang to 1.20.
- Updated all dependencies to their latest version.
- Use unpriviliged
appuser
with UID/GID 4200 instead of nobody user. - Use
sigs.k8s.io/yaml
instead ofgithub.com/ghodss/yaml
.
- Updated all dependencies to their latest version.
- Update resources using
Patch
instead ofUpdate
. - Use RFC3339 time format for logs.
- Update AbsencePrometheusRule if only labels have changed.
- Do not requeue resources that do not exist.
- Do not log error if corresponding AbsencePrometheusRule can not be retrieved or does not exist during clean up.
- Use debug log level for less important log messages.
debug
flag.
- Carry over
support_group
labels from original alert rules.
- Updated all dependencies to their latest version.
- Add missing Kubebuilder annotations.
- Logging during metric expression parsing.
keep-labels
flag parsing.- Prevent getting stuck during clean up of orphaned absence alert rules when the corresponding AbsencePrometheusRule doesn't exist.
- Skip metrics that match against the internal
__name__
label and useabsent
function.
- Support for determining
tier
label heuristically has been deprecated and will be removed in a later version.
- Use Kubebuilder for scaffolding.
absent-metrics-operator/updated-at
annotation to operator generatedPrometheusRule
resources which specifies the time (UTC) at which this resource was updated by the operator.
- Update Ginkgo testing framework to v2.
- Updated all dependencies to their latest version.
- Updated all dependencies to their latest version.
- Updated Go to
1.17
and all dependencies to their latest version.
- Clean up of PrometheusRule resources for which the operator is disabled.
- Clean up orphaned gauge metrics.
- Manual maintenance task.
- Delete timeseries concerning
PrometheusRules
that no longer exist.
- A bug that was introduced in the previous release.
- Prevent superfluous processing if the resource doesn't have any alert rules.
context
label to absence alerts.
playbook
label from absence alerts.
- Parse vector selectors that use label matching against the internal
__name__
label.
- Operator can be disabled for a specific alert rule.
playbook
label to absence alerts.keep-labels
flag for specifying which labels to carry over from alert rules.
- Initial release.