[openstack-helm] Get rid of cephfs and rbd provisioners
Hi everyone,
I would like to suggest getting rid of cephfs and rbd provisioners. They have been retired and have not been maintained for about 2.5 years now [1]. I believe the CSI approach is what all users rely on nowadays and we can safely remove them.
The trigger for this suggestion is that we are currently experiencing issues while trying to switch cephfs provisioner to Ubuntu Focal and fixing this is just wasting time. [2] Stephen spent some time debugging the issues and can give more details if needed.
What do you think?
[1] https://github.com/kubernetes-retired/external-storage/tree/master/ceph [2] https://review.opendev.org/c/openstack/openstack-helm-infra/+/872976
Hi Vladimir,
I agree. I also think we should stop maintaining the CSI provisioner chart and simply deploy the one provided by the Ceph CSI team
Less code we maintain, the better.
Thanks Mohammed
On Thu, Mar 2, 2023 at 10:13 PM Vladimir Kozhukalov kozhukalov@gmail.com wrote:
Hi everyone,
I would like to suggest getting rid of cephfs and rbd provisioners. They have been retired and have not been maintained for about 2.5 years now [1]. I believe the CSI approach is what all users rely on nowadays and we can safely remove them.
The trigger for this suggestion is that we are currently experiencing issues while trying to switch cephfs provisioner to Ubuntu Focal and fixing this is just wasting time. [2] Stephen spent some time debugging the issues and can give more details if needed.
What do you think?
[1] https://github.com/kubernetes-retired/external-storage/tree/master/ceph [2] https://review.opendev.org/c/openstack/openstack-helm-infra/+/872976 -- Best regards, Kozhukalov Vladimir
I agree as well. I have been working to update the Ceph client images to Quincy and Focal to match the Ceph daemons that have already been updated, and the CephFS provisioner is proving difficult. It is outdated and incompatible with Python 3, but newer librados packages are only built for Python 3.
If we want to keep the old provisioners around, the path that makes the most sense is to update them to be compatible with more modern frameworks and libraries, but personally, I don't see a need. I think pretty much everyone has moved to CSI, and anyone that hasn't probably should. I am in favor of removing the outdated provisioners.
Steve On 3/2/2023 3:22:20 PM, Mohammed Naser mnaser@vexxhost.com wrote: Hi Vladimir,
I agree. I also think we should stop maintaining the CSI provisioner chart and simply deploy the one provided by the Ceph CSI team
Less code we maintain, the better.
Thanks Mohammed
On Thu, Mar 2, 2023 at 10:13 PM Vladimir Kozhukalov <kozhukalov@gmail.com [mailto:kozhukalov@gmail.com]> wrote:
Hi everyone,
I would like to suggest getting rid of cephfs and rbd provisioners. They have been retired and have not been maintained for about 2.5 years now [1]. I believe the CSI approach is what all users rely on nowadays and we can safely remove them.
The trigger for this suggestion is that we are currently experiencing issues while trying to switch cephfs provisioner to Ubuntu Focal and fixing this is just wasting time. [2] Stephen spent some time debugging the issues and can give more details if needed.
What do you think? [1] https://github.com/kubernetes-retired/external-storage/tree/master/ceph [https://github.com/kubernetes-retired/external-storage/tree/master/ceph]
[2] https://review.opendev.org/c/openstack/openstack-helm-infra/+/872976 [https://review.opendev.org/c/openstack/openstack-helm-infra/+/872976] --
Best regards, Kozhukalov Vladimir
--
Mohammed Naser VEXXHOST, Inc.
Guys,
Thank you for your thoughts. I appreciate it. Looks like we all agree about removal of old style provisioners. I'll prepare a PS for this.
Also thanks for the good idea to switch to the upstream charts for CSI Ceph provisioners [1]. Let's do this as a separate PS.
[1] https://github.com/ceph/ceph-csi/tree/devel/charts
On Mon, Mar 6, 2023 at 6:40 PM Steve Taylor steveftaylor@gmail.com wrote:
I agree as well. I have been working to update the Ceph client images to Quincy and Focal to match the Ceph daemons that have already been updated, and the CephFS provisioner is proving difficult. It is outdated and incompatible with Python 3, but newer librados packages are only built for Python 3.
If we want to keep the old provisioners around, the path that makes the most sense is to update them to be compatible with more modern frameworks and libraries, but personally, I don't see a need. I think pretty much everyone has moved to CSI, and anyone that hasn't probably should. I am in favor of removing the outdated provisioners.
Steve
On 3/2/2023 3:22:20 PM, Mohammed Naser mnaser@vexxhost.com wrote: Hi Vladimir,
I agree. I also think we should stop maintaining the CSI provisioner chart and simply deploy the one provided by the Ceph CSI team
Less code we maintain, the better.
Thanks Mohammed
On Thu, Mar 2, 2023 at 10:13 PM Vladimir Kozhukalov kozhukalov@gmail.com wrote:
Hi everyone,
I would like to suggest getting rid of cephfs and rbd provisioners. They have been retired and have not been maintained for about 2.5 years now [1]. I believe the CSI approach is what all users rely on nowadays and we can safely remove them.
The trigger for this suggestion is that we are currently experiencing issues while trying to switch cephfs provisioner to Ubuntu Focal and fixing this is just wasting time. [2] Stephen spent some time debugging the issues and can give more details if needed.
What do you think?
[1] https://github.com/kubernetes-retired/external-storage/tree/master/ceph [2] https://review.opendev.org/c/openstack/openstack-helm-infra/+/872976 -- Best regards, Kozhukalov Vladimir
-- Mohammed Naser VEXXHOST, Inc.
participants (3)
-
Mohammed Naser
-
Steve Taylor
-
Vladimir Kozhukalov