Releases

Additional features we're working on:

• Cortex Alert Manager :
  • scalability & HA consistency improvements for opni-alerting

v0.10.1 (TBA)

• Gateway/Agent
  • Add ping endpoint to gateway
  • Add rate limiting to streaming connections
  • Auto update of agent
• Logging
  • Bug fixes to controlplane logs and events

v0.10.0 (June 6, 2023)

• Metrics
  • Use OTEL collector for metrics collection (greatly reduced agent footprint)

v0.9.2 (April 26, 2023)

• AIOps enhancements/fixes
  • Update AI services to now give clearer statuses of the Deep Learning model when training. (#1272)
  • Fix bug where updating a watchlist by simply removing a deployment would cause training status to hang. (#1344)
• Alerting enhancements/fixes
  • Resolve internal data race with internal alerting conditions (#1174)
  • Further improved performance of alarm status listing (#1022)
• Monitoring fixes:
  • Fixed an issue where an agent reconnecting could cause errors scraping gateway metrics (#1295)
• CLI fixes:
  • Fixed bash completion not working correctly (#1299)

v0.9.1 (April 6, 2023)

• AIOps enhancements/fixes
  • Update AI Services to now use Nats Jetstream KV storage (#1146)
  • Remove "Enable GPU Services" button from Opni Admin Dashboard. (#1147)
  • Update AIOps gateway plugin to update Nats Jetstream kv when model is submitted for training. (#1148)
  • Update AI Services to launch model training upon startup of service (#1233)

v0.9.0 (March 16, 2023)

• Alerting enhancements

  • Metrics based alarm incidents are now tracked in the alerting timeline

• Monitoring enhancements

  • Opni internal metrics containing cluster names are now stored per-tenant instead of in the local cluster's tenant
  • Upstream Opni local agent identification

• Logging enhancements

  • Move to using open-telemetry-collector to collect logs
  • Send logs to the central cluster using gRPC
  • Store cluster id to name mapping for better visibility

• AIOps enhancements/fixes

  • Update Opensearch query to get log count per deployment based on updated field names from open-telemetry collector ( #1181 )
  • Update Opensearch query to use updated field names from open-telemetry-collector to fetch logs from Opensearch for training Deep Learning model. ( #19 )

• UI

  • Fixed a bug that was preventing the Alert timeline to fully render
  • Added a friendly name field to the Manual Installation section of Add Agent to coincide with a chart update. (#1177)
  • Tagging Agents in the table with an icon which indicates if the agent is local or not. (#1166)

v0.8.3 (March 2, 2023)

• AIOps Enhancements/Fixes

  • Optimize query for filtering out anomalous keywords within training controller. ( #18 )
  • implement streaming data-loader for log anomaly detection model training to reduce memory pressure from large training dataset. ( pr40, pr17)

• UI Enhancements/Fixes

  • Fix an issue where the agents page crashed when an agent no longer exists for a role that was applied to the agent ( #1109 )
  • Fixed Monitoring capability metrics not showing up ( #1111 )
  • Changing how we add agents via the UI to improve reliability ( #965 )
  • Switched to using the new alarm API. ( pr118 )

v0.8.2 (February 16, 2023)

• Monitoring enhancements
  • metrics admin CLI can list / filter & aggregate cortex rules
    • includes listing & querying stateful information about alerting rules
  • metrics admin CLI can create,read, update or delete standard prometheus rule group files to clusters
  • An initial cluster name can be set during bootstrap of new agents by setting the opni-agent.friendlyName helm value. When a cluster name is set in the UI, this flag will be added to the copyable helm install command.
• Minor update to ingest plugin to prepare for Opni preprocessing service
• Notable dependency updates
  • go 1.20
  • cortex to v1.14.1
  • go-plugin to v1.4.8
  • nats-server to v2.9.12
  • alertmanager to v0.25.0
  • prometheus to v0.41.0
  • opentelemetry to v1.12.0
  • grpc to v1.52.1
  • kubernetes sdk to v0.26.1
  • controller-runtime to v0.14.4
  • cert-manager to v1.10.2
  • gin to v1.8.2
• UI Enhancements/Fixes
  • Updating the tooltip in the alerting overview page to improve accuracy (#1006)
  • Fixed a bug where we didn't show all log templates in the Opensearch plugin. Also added a count to the top of the same table. (pr 106)
  • Fixed a bug where you couldn't update the s3 endpoint in the monitoring config (#1009)
  • Fixed a bug where you couldn't edit PagerDuty alarm endpoints (#958)
  • Improved the consistency of Cluster/Agent name and id labeling (#1048)
  • Improved chart labeling on the insights page of the opensearch dashboard (#1046)
  • Only show relevant storage options when selecting the 'Highly Available' Monitoring option
  • Compensate for an overzealous 'no changes to apply' error the Alerting backend emits when installing Alerting.
• AIOps Enhancements/Fixes
  • Improved pre-trained DRAIN models with more ground truth (pr16)
  • Added test coverage to 64% to the deep-learning model service (pr38)
  • Fixed the NotFoundError during downloading training data from Opensearch(pr39)
  • Updated Opensearch query to filter out anomaly keywords in training data (pr16)
  • Bump PyTorch version to latest stable version (pr1035)
• Alerting enhancements
  • scalability optimizations; reduced memory footprint for the Alerting backend
  • Improved fault tolerance for the Alerting backend notification system

v0.8.1 (December 18, 2022)

Bugfix release:

• Add retry for fetching client certs
• Fix agent crash on startup

v0.8.0 (December 16, 2022)

Introduction of Opni AI workload training module:

• Pre-req: NVIDIA GPU is enabled in cluster Opni is installed in
• Users can select deployments are important to them in the Opni Admin UI -> a model will be trained for the user and logs belonging to those deployments in the future will be given a "Normal" or "Anomalous" label
• Users can ingest these insights in the Opni plugin in their Opensearch Dashboards (enabled when Opni Logging backend is setup)

Alerting enhancements

• New PagerDuty endpoint for receiving notifications
• Clone operation for cloning alarms to any target cluster(s)
• Alarms for when downstream agent capabilities are unhealthy
• Alarms for tracking the health of the opni monitoring backend health
• Alarms based on prometheus queries
• Alarms for tracking the states of kubernetes objects

Opensearch changes:

• Opensearch version updated to 2.4.0
• Dataprepper version updated to 2.0.1
• Reconcilers now user TLC client cert to authenticate to the Opensearch API

v0.7.0 (November 23, 2022)

Introduction of Opni Alerting:

• Slack, Email Endpoint for receiving notifications
• Alarms on agent connection status
• Overview for breached alarms