Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

New Upstream Components

Significant Upstream Version Updates

...

Fixes

  • Resilience of some critical system services (e.g. ntpd, kubelet, etc.) improved. In case of failure services recovers by themselves.
  • Elasticsearch cluster made to more robust. Also rotation set for java heap dump files.
  • Security corrections (problems discovered by OpenScap Security Scan, and corrections for CVEs in upstream components).
  • Pod eviction improved by defining memory thresholds.

Significant Upstream Version Updates

  • CPU Pooler 0.3.0
  • DANM 4.1.0
  • Elasticsearch 7.4.2
  • Etcd 3.4.3
  • Fluentd 1.7.4
  • Helm 2.15.2
  • KubeDNS 1.15.7
  • Kubernetes 1.16.2
  • Prometheus 2.13.0
  • Swift 2.23.1

...

Currently REC only supports bare metal installation. In place upgrades are not supported. The upgrade procedure is the same as the inital initial installation procedure documented in the REC Installation Guide.

...

Known Limitations, Issues and Workarounds

  • Fill in list here prior to releaseA fix to CPUPooler for allocation from both shared & exclusive CPU pools: https://gerrit.akraino.org/r/c/ta/caas-cpupooler/+/2015 has been been merged in Gerrit after the release candidate build 237 was built and is not part of REC Release 2. This is a minor correction which is not currently required by the RIC.
  • An improvement for Kubernetes node HA: https://gerrit.akraino.org/r/c/ta/caas-kubernetes/+/1961 has been pushed to Gerrit but has not been merged. It is not included in release 2. In some cases when a Kubernetes node becomes unschedulable, it may need manual intervention to make node schedulable again. Kubernetes functions as expected and simply requires an admin to uncordon the node when it has recovered.

References