Releases: solarwinds/swi-k8s-opentelemetry-collector
Releases · solarwinds/swi-k8s-opentelemetry-collector
swo-k8s-collector-3.2.0-alpha.3
- Added option to enable opencost metrics
- Fine tuning default node-collector values for better performance
swo-k8s-collector-3.1.0
Added
- Added network monitoring for Linux nodes which is based on eBPF, so it does not require service mesh to be deployed.
- It is disabled by default, to enable it set
ebpfNetworkMonitoring.enabled: true
invalues.yaml
- It is based on opentelemetry-ebpf
- For scaling see
numIngestShards
,numMatchingShards
andnumAggregationShards
invalues.yaml
- See exported_metrics.md for list of metrics
- It is disabled by default, to enable it set
- Automatic discovery and scraping of prometheus endpoints on pods. Driven by
otel.metrics.autodiscovery.prometheusEndpoints.enabled
option invalues.yaml
, by default enabled (Fargate not yet supported).- In case
otel.metrics.autodiscovery.prometheusEndpoints.enabled
is set totrue
(which is by default)extra_scrape_metrics
is ignored as there is high chance that those metrics will be collected two times. You can override this behavior by by settingforce_extra_scrape_metrics
to true.
- In case
- Added option to set
imagePullSecrets
invalues.yaml
- Added option to configure
terminationGracePeriodSeconds
defaulting to 10 minutes, so that it is guaranteed that collector process whole pipeline - Added option to configure
sending_queue
- Added option to offload in metrics collector
sending_queue
to storage, reducing memory requirement for the collector
- Added option to offload in metrics collector
- Added option to configure
retry_on_failure
- default for initial_interval is now
10s
(previously was5s
) avoiding unnecessary retries when backend takes time to respond
- default for initial_interval is now
Changed
- Prometheus is no longer needed for kubernetes monitoring, therefore it is no longer deployed. By default, k8s collector does not scrape anything from Prometheus.
- If there was a Prometheus deployed by the previous version (using setting `prometheus.enabled``), it will be automatically removed.
- Setting
otel.metrics.prometheus.url
is still available, but is valid only in combination withotel.metrics.extra_scrape_metrics
. - If
otel.metrics.prometheus.url
is empty,otel.metrics.prometheus_check
is ignored to not fail on missing Prometheus. - Note that
otel.metrics.extra_scrape_metrics
is deprecated option and will be removed in future versions. - Node metrics are now scaped from node-collector daemonset
- Decreased the default batch size for metrics, logs and events sent to OTEL endpoint to 512 to avoid too big messages
- Upgraded SWO Agent image to
v2.6.28
0.8.12
swo-k8s-collector-3.2.0-alpha.1
Added
- Added ARM64 support
- Upgraded OTEL collector image to
0.8.12
(see Release notes)
- Upgraded OTEL collector image to
0.8.11
swo-k8s-collector-3.1.0-rc.6
SolarWinds Kubernetes Integration
swo-k8s-collector-3.1.0-rc.1
Changed
sending_queue
.queue_size
changed from1000
to200
, decreasing amount of memory it can take- Upgraded OTEL collector image to
0.8.10
(see Release notes) - Upgraded SWO Agent image to
v2.6.28
Added
- Added option to configure terminationGracePeriodSeconds defaulting to 10 minutes, so that it is guaranteed that collector process whole pipeline
- Added option to offload sending_queue to storage, reducing memory requirement for the collector
- Added option to configure sending_queue
- Added option to configure retry_on_failure
- default for initial_interval is now
10s
(previously was5s
) avoiding unnecessary retries when backend takes time to respond
- default for initial_interval is now
Fixed
- ebpf monitoring: Added necessary init containers making sure that all components start in the right order
0.8.10
What's Changed
- NH-66305: Optimizing amount of stored data in k8sattributes by @gantrior in #449
- Bump library/golang from 1.21.3-bookworm to 1.21.4-bookworm in /build/docker by @dependabot in #460
Full Changelog: 0.8.9...0.8.10
swo-k8s-collector-3.1.0-alpha.2
Added option to set imagePullSecrets
swo-k8s-collector-3.0.0
Changed
- Prometheus is no longer needed for kubernetes monitoring, therefore it is no longer deployed. By default, k8s collector does not scrape anything from Prometheus.
- If there was a Prometheus deployed by the previous version (using setting `prometheus.enabled``), it will be automatically removed.
- Setting
otel.metrics.prometheus.url
is still available, but is valid only in combination withotel.metrics.extra_scrape_metrics
. - If
otel.metrics.prometheus.url
is empty, `otel.metrics.prometheus_check`` is ignored to not fail on missing Prometheus. - Note that
otel.metrics.extra_scrape_metrics
is deprecated option and will be removed in future versions.
- Decreased the default batch size for metrics, logs and events sent to OTEL endpoint to 512 to avoid too big messages
- Upgraded OTEL collector image to
0.8.8
(see Release notes)
Added
- Added network monitoring for Linux nodes which is based on eBPF, so it does not require service mesh to be deployed.
- It is disabled by default, to enable it set
ebpfNetworkMonitoring.enabled: true
invalues.yaml
- It is based on opentelemetry-ebpf
- See exported_metrics.md for list of metrics
- It is disabled by default, to enable it set