Unsolved
36 Posts
0
1111
petition to support rancher
I would like to recommend support for rancher/RKE. I have been using the isilon CSI in my clusters and have not come across any major road blocks. With the release of 1.2.0 of the isilon CSI and the move to helm3, the installation is pretty straight forward. The only change I made for 1.2.0 was the following in the verify.kubernetes script. RKE uses the term controlplane instead of master. So this is the change I made.
# Determine the nodes
MINION_NODES=$(kubectl get nodes -o wide | grep -v -e master -e INTERNAL -e controlplane | awk ' { print $6; }')
MASTER_NODES=$(kubectl get nodes -o wide | awk ' /master|controlplane/{ print $6; }')
Maybe this could be a 1.2.1 release. It's very minor and makes it easy to deploy to RKE.
On another note, the recaptha is getting ridiculous. I spent more time clicking to add a post than I do writing the post.
Flo_csI
166 Posts
0
June 24th, 2020 02:00
Hi @cocampfe,
The PowerMax driver has an option to skip the checks https://github.com/dell/csi-powermax/blob/master/helm/install.powermax#L18
I will request we get it for every drivers.
+1 on the reCaptcha.... but there is nothing I can do about it
cocampbe
36 Posts
0
June 24th, 2020 10:00
I don't know that I would skip the checks. But that assumes we are talking about the same checks. I think it is valid to check for the feature gates. EMC just needs to keep up to speed on which gates need to be added.
lol on the recaptcha. I think I have clicked about 100 fire hydrants, cars, and crosswalks since joining the discussions.
cocampbe
36 Posts
0
June 26th, 2020 05:00
@Flo_csI I've been thinking about this over the last few days. I just wanted to share some thoughts about the isilon csi.
Maybe the install and uninstall scripts should be removed. I was thinking that the the verify.kubernetes script should stay, and that the install guide could state to use that to verify you have the feature gates set. Then show the helm commands to install and uninstall the csi in the docs. This comes to mind, because I have toyed with the idea of using fluxcd to deploy the csi. We did it in test env and it worked well. And I have also been thinking about how i could use kustomize for all the different clusters we have. I like to change the prefix per cluster so I know which export relates to a cluster.
I guess at some level, I think people should be learning helm. The script kind of abstracts that away. That's not a bad thing. In this whole new world of k8s I think it's important to get people used to using all these tools.