Skip to content
Pre-release
Pre-release

@kmova kmova released this Nov 19, 2018

Getting Started

Prerequisite to install

  • Kubernetes 1.9.7+ is installed
  • Make sure that you run the below installation steps with cluster admin context. The installation will involve creating a new Service Account and assigning to OpenEBS components.
  • Make sure iSCSI Initiator is installed on the Kubernetes nodes.
  • NDM helps in discovering the devices attached to Kubernetes nodes, which can be used to create storage pools. If you like to exclude some of the disks from getting discovered, update the filters on NDM to exclude paths before installing OpenEBS.

Using kubectl

kubectl apply -f https://openebs.github.io/charts/openebs-operator-0.7.2.yaml

Using helm stable charts

helm repo update
helm install  --namespace openebs --name openebs stable/openebs

Sample Storage Pool Claims, Storage Class and PVC configurations to make use of new features can be found here: Sample YAMLs

For more details refer to the documentation at: https://docs.openebs.io/

Change Summary

Minor enhancements

  • Support for clearing space used by the Jiva Replica after the Jiva Volume is deleted. The space reclaim is done by scheduling a Kubernetes Job. Note: If your cluster setup requires lots of OpenEBS volumes to be created and deleted, you will need to setup a cron job to clear the completed jobs. This is required, till Kubernetes feature TTL Mechanism for Finished Jobs is supported.
  • Support for a storage policy that can disable the Jiva Volume Space reclaim. Please add the following to your StorageClasses, if you would like to disable Jiva Volume Space reclaim (or in other words - retain the volume data post PVC deletion).
      cas.openebs.io/config: |
        - name: RetainReplicaData
          enabled: true  
    
  • Support for a storage policy to allow scheduling the Jiva Target Pods on the same node as the Application Pod. This feature makes use of the Kubernetes Pod Affinity feature that is dependent on the Pod labels. You will need to add the following label to both Application and PVC.
    labels:
      openebs.io/target-affinity: <application-unique-label>
    

Bug Fixes

  • Fixed an issue where internal snapshots created for rebuilding were getting deleted in case of errors in opening or holding the snapshots.
  • Fixed an issue in sending cStor volume metrics to Prometheus like uptime, read/write block count and read/write time(ns)

Detailed release notes are maintained in Project Tracker Wiki.

Limitations

  • Jiva target to Replica message protocol has been enhanced to handle the write errors. This change in the data exchanges causes the older replicas to be incompatible with the newer target and vice versa. The upgrade involves shutting down all the replicas before launching them with the new version. Since the volume requires the target and at least 2 replicas to be online, chances of volumes getting into the read-only state during upgrade are higher. A manual intervention will be required to recover the volume.
  • For OpenEBS volumes configured with more than 1 replica, at least more than half of the replicas should be online for the Volume to allow Read and Write. In the upcoming releases, with cStor data engine, Volumes can be allowed to Read/Write when there is at least one replica in the ready state.
  • This release contains a preview support for cloning an OpenEBS Volume from a snapshot. This feature only supports single replica for a cloned volume, which is intended to be used for temporarily spinning up a new application pod for recovering lost data from the previous snapshot.
  • While testing for different platforms, with a three-node/replica OpenEBS volume and shutting down one of the three nodes, there was an intermittent case where one of the 2 remaining replicas also had to be restarted.
  • The OpenEBS target (controller) pod depends on the Kubernetes node tolerations to reschedule the pod in the event of node failure. For this feature to work, TaintNodesByCondition alpha feature must be enabled in Kubernetes. In a scenario where OpenEBS target (controller) is not rescheduled or is back to running within 120 seconds, the volume gets into a read-only state and a manual intervention is required to make the volume as read-write.
  • The current version of OpenEBS volumes are not optimized for performance sensitive applications.

For a more comprehensive list of open issues uncovered through e2e, please refer to open issues.

Assets 2
You can’t perform that action at this time.