Chapter 3. Enhancements

This section describes major enhancements introduced in Red Hat OpenShift Container Storage 4.7.

Better indication for successful upgrade of OpenShift Container Storage

Previously, it was difficult to determine if an OpenShift Container Storage upgrade had finished successfully. In some cases, the console would report everything was fine while in reality some components had not been upgraded to their new container images. With this update, the StorageCluster now checks and reports the running container images for all of its managed components, which aids in troubleshooting upgrade scenarios

OSD Pod Disruption Budgets redesign

Previously, the OpenShift Container Storage Pod Disruption Budgets (PDBs) by default have a minUnavailable=0 and only allow rebooting of OSDs on a single node at a time. This caused the OCP console to constantly show a warning about nodes not being able to restart. With this update, OSD PDBs have the following redesigns:

  • There is one OSD PDB in the beginning. This allows only one OSD to go down at any time.
  • Once an OSD goes down, its failure domain is determined and any blocking OSD PDBs are created for other failure domains.
  • The original OSD PDB created is deleted. As a result, all the OSDs can go down in the failure domain.

With the new design, users can drain multiple nodes in the same failure domain.

Update RGW address in the external mode

With this update, you can change the RGW address without affecting the Multicloud Object Gateway’s(MCG) work if the MCG is configured with an RGW backingstore in the external mode.

Free space on RGW

Earlier, NooBaa bucket showed the storage capacity of 1PiB for all the buckets and did not show the free space on RGW. With this enhancement, the storage capacity of the Red Hat Ceph cluster is exported in the status field and now the NooBaa operator listens to the changes to this status field and updates the available capacity of every RGW based backingstore.

Allow the configuration of the Service monitor port to differ from the default ceph-mgr Prometheus port for external mode

With this enhancement, if the external Red Hat Ceph cluster is configured with a ceph-mgr Prometheus module listening on the non-default port (9283) then OpenShift Container Storage now can connect and consume those metrics, that is, the OpenShift Container Storage now accepts any monitoring port.

ocs-operator accepts non-default ports for monitoring services for external mode

Previously, there were no provisions in the ocs-operator to pass on a monitoring Prometheus service port other than the default port 9283. This made the port unusable for monitoring services. With this update, the ocs-operator has been enabled to accept and propagate non-default monitoring ports from external cluster JSON input and monitoring services work as expected.

Use an existing secret to create a new backingstore

With this enhancement, a new backingstore can be created through the Multicloud Object Gateway CLI using an existing secret.

Prioritize creating new OSD deployments over updating existing OSD deployments

Previously, for OSDs on Persistent Volume Claims, Rook implicitly favored updating existing OSDs before creating new OSDs resulting in new capacity not being added to the cluster until the end of the OSD reconcile. With this enhancement, the cluster now favors scaling up OSDs over updating existing OSDs to make new capacity available as soon as storage has been provisioned and this also reduces the reconcile time to 5-10 minutes from 15 while scaling up the number of OSDs in the cluster.

Public route for RGW

With this update, the OpenShift Container Storage operator now creates a route for the Red Hat Ceph Storage’s RADOS Object Gateway (RGW) service.

OpenShift Container Storage deployed on ROKS without credentials in the IBM cloud setup

With this update, for the ease of installation in the IBM cloud setup when OpenShift Container Storage is deployed on ROKS and no credentials are provided for using ROKS as the default backingstore, a PV pool default BackingStore is created.

Prometheus alert for OSD restart

This enhancement adds a Prometheus alert to notify if an OpenShift Container Storage OSD restarts more than 5 times in 5 minutes. The alert message is as follows:

 Storage daemon osd.x has restarted 5 times in the last 5 minutes. Please check the pod events or ceph status to find out the cause.

where, x represents the OSD number.

System alerts for namespace buckets

With the release of the Red Hat OpenShift Container Storage 4.7, system alerts have been added for the namespace buckets and resources to get a better understanding of the current state of the system.

Log messages printing to the noobaa-endpoint pod log

Earlier, log messages were printed to the noobaa-endpoint pod log even if the debug option was not set. With this release, the log messages are printed to the noobaa-endpoint pod log only when the debug option is set.