v2.13.0 - 12/Nov/2026

Global Tolerations and Scheduling Precedence: Added global.suseObservability.tolerations to apply tolerations to every SUSE® Observability workload. Affinity and tolerations now follow one precedence order, with the most specific setting winning. This changes existing behavior, see the upgrade instructions for v2.13.0. For details, see Configure Kubernetes affinities and tolerations.

Granular Pod Metadata: Added support for custom pod labels and annotations on HBase workloads (including the integrity CronJob) and the KafkaUp operator, as well as custom pod annotations for vmagent and anomaly detection workloads. Component-level configurations override shared defaults while safely preserving mandatory selector labels and configuration checksums.

AD FS OIDC Compatibility: Added the stackstate.authentication.oidc.omitClientNameParameter Helm value to support Active Directory Federation Services (AD FS) OIDC implementations that reject the ?client_name=StsOidcClient query parameter.

LDAP Certificate Validation Bypass: Added the stackstate.authentication.ldap.ssl.skipCertificateValidation Helm value to encrypt the LDAP connection without validating the server certificate. Use it only temporarily, until the CA certificate of the LDAP server is available.

Monitor Template Expressions: Placeholders in urnTemplate and titleTemplate can now contain CEL expressions that use the time series labels.

Dynamic Kafka JVM Sizing: Kafka now automatically calculates and sets its JVM heap size to 50% of the effective container memory request, while respecting any explicitly configured kafka.heapOpts overrides.

Node Drain & Maintenance Optimization: Improved node draining workflows during cluster upgrades by permitting the eviction of unhealthy platform and agent pods while strictly honoring Pod Disruption Budgets (PDBs) for healthy instances.

Trace Tail-Sampling Configuration: Added configurable settings for tail-sampling wait times, internal buffer capacities, and decision caching to better manage trace data volume.

Resource Naming Standardization:

HA Profile Retention: The 4000-ha profile now defaults to the standard log retention policy.

Tracing Query Duplication: Fixed a bug where the traces perspective loaded data twice on page load, occasionally rendering incorrect query results.

Authentication Race Condition: Fixed intermittent login failures and infinite redirect loops triggered by concurrent authentication requests during browser initialization.

vmagent Buffer Protection: Resolved an issue where vmagent exhausted available disk space and halted execution during periods of backpressure when VictoriaMetrics lagged in metric ingestion. The vmagent buffer is now capped at 80% of volume capacity to account for filesystem overhead, allowing metric collection to continue smoothly while VictoriaMetrics catches up.

HBase & HDFS Resilience: Fixed an issue where HDFS dropped HBase region Write-Ahead Log (WAL) blocks during simultaneous region downtime and DataNode restarts.

Backup PVC Exhaustion: Resolved an issue where the configuration backup Persistent Volume Claim (PVC) ran out of disk space during regular operation.

Topology Sync Stalls: Fixed an issue where the topology synchronization pod became stuck due to an unhandled name: must not be empty validation failure.

Helm Chart Packaging: Fixed a packaging issue in the SUSE Observability Agent Helm chart to guarantee all required subcharts are bundled into the release archive.