Previously, the 'endpoints' and 'service' on the 'projects/namespaces' had to be created to use Red Hat Gluster Storage persistent volumes. With this release, when the volumes are dynamically provisioned by the gluster provisioner, the subjected 'endpoint' and 'service' are auto created and the user will be able to use these persistent volumes.
Description of problem:
== This is not a Heketi bug. This is registered here to track the upstream GlusterFS adapter bug. We should move this to an 'rhgs-ose' component when available ==
We thought at first that we could create a pool of GlusterFS PVs available for any namespace PVC to bind to. The issue is that GlusterFS adapter requires the PV to refer to an Endpoint, and endpoints are per namespace only, even though PVs are cluster wide.
The usability of this model would require the following:
Admin:
First, create a set of GlusterFS volumes (no PVs yet)
User:
Create a namespace
Create a PVC before creating the PV
Create Endpoints object in namespace
Create PV pointing to the endpoints name in the namespace
Deploy Pod
Instead we would like to suggest that the PV contain all the information needed to mount a GlusterFS volume, since PVs are available cluster wide. This way a PV would not need to point to an Endpoint, since they are only available per namespace.
Tracker for https://github.com/kubernetes/kubernetes/issues/28791
Hi Humble,
I have made minor edits to the doc text to be included in the errata. Let me know if it looks ok and sign off if you have no review comments.
Thanks
Comment 10Humble Chirammal
2017-01-09 08:21:44 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://rhn.redhat.com/errata/RHEA-2017-0148.html
Description of problem: == This is not a Heketi bug. This is registered here to track the upstream GlusterFS adapter bug. We should move this to an 'rhgs-ose' component when available == We thought at first that we could create a pool of GlusterFS PVs available for any namespace PVC to bind to. The issue is that GlusterFS adapter requires the PV to refer to an Endpoint, and endpoints are per namespace only, even though PVs are cluster wide. The usability of this model would require the following: Admin: First, create a set of GlusterFS volumes (no PVs yet) User: Create a namespace Create a PVC before creating the PV Create Endpoints object in namespace Create PV pointing to the endpoints name in the namespace Deploy Pod Instead we would like to suggest that the PV contain all the information needed to mount a GlusterFS volume, since PVs are available cluster wide. This way a PV would not need to point to an Endpoint, since they are only available per namespace. Tracker for https://github.com/kubernetes/kubernetes/issues/28791