Skip to content

v0.23.0-rc3

Pre-release
Pre-release

Choose a tag to compare

@mickmis mickmis released this 11 Aug 11:49
· 37 commits to master since this release
590e569

Release Notes for v0.23.0-rc3

⚠️ This release candidate release contains a known issue when deploying to AWS due to the storage class used. ⚠️

Mandatory migration tasks

N/A

Optional migration tasks

  • enable_opentelemetry has been split into two flags: enable_metrics and enable_tracing
  • A new pair of migrations for CockroachDB has been added for RID (4.1.0) and SCD (3.4.0) to remove useless indexes.
    • The DSS will work regardless of the migration status.
    • This migration will improve performance. If that is not the case for a specific scenario that was missed, please downgrade to the previous version and open an issue.
    • These indexes are still useful for Yugabyte, so no such migration exists for it.
    • Details about migrations are available there
  • The 3.4.1 (or 1.1.1 for Yugabyte) migration for SCD is mandatory if you plan to use the new SCD hash lock option.
    • This migration can be applied without any impact if you do not plan to use it.

Important information

  • Fixed a bug where the evict command ignored entries without a locality. If your DSS instance does not have a locality set, the next evict run may be slow while it processes the backlog of old entries.

  • Fixed a bug where Helm charts and Tanka files didn't actually perform any actions via the evict command when run via cron jobs (because no locality was set and no delete flag was specified). If you have a large number of entries, the next run may be slow while it processes the backlog of old entries.

  • AWS load balancer names are no longer enforced by Helm charts or Tanka files. Existing clusters will retain their current names, while new ones will use names automatically generated by AWS.

  • Server timeout flag has been renamed from server timeout to server_timeout.

  • The duration field, previously recorded as a string containing the unit, has been changed to duration_ms, a float representing the duration in milliseconds, rounded to 0.01.

  • Grafana version deployed by Helm charts or Tanka files have been upgraded to 13.0. Ensure to read grafana changelog based on your current version.

  • The Google Kubernetes Engine (GKE) node disk size has been increased from 15GB to 25GB to prevent issues with saturated system disks.

    • You will need to apply the Terraform state once to recreate the node pool.
    • This process will take some time because it is performed in a non-disruptive rolling update, draining and moving pods across nodes during the upgrade.
    • Please plan accordingly and test the rollout on a non-production cluster first.

Minimal database schema version

Schema CockroachDB Yugabyte
AUX 1.1.0 1.1.0
RID 4.0.0 1.0.1
SCD 3.3.0 1.1.0

Validation

USS qualifier versions: v0.31.0
Test configuration and report: See attachment.

What's Changed

New Contributors

Full Changelog: interuss/dss/v0.23.0-rc2...interuss/dss/v0.23.0-rc3