Feature Request: make Ubuntu template to support LVM

[mbelchin]

Hi team,

this is a question but also a feature request.

Is there any ubuntu template we can use with CSE that support LVM ?

This is our case:

On creating new clusters with CSE we're using this template:

ubuntu-16.04-server-cloudimg-amd64-k8s
Ubuntu 16.04 Docker 18.03.0~ce Kubernetes 1.10.1 weave 2.3.0

After that we needed to add more disk space to one of the nodes and we discovered that this image does not support LVM.

LVM is an abstraction layer for the storage on the machine, you can add more disks and then extend the volume groups and the logical volumes. It takes immediately effect and no reboot is needed, plus you don't need to touch partition table at all.

Would be great if the Ubuntu 16.04 template could support LVM.

[rocknes]

Hi mbelchin,

We don't have any plans to support templates that can leverage LVM in near future. However please feel free to modify the auto generated CSE templates,

1. deploy a vApp using the CSE generated templates
2. install LVM related software on it.
3. recapture them back as templates.

You can replace (by renaming) the CSE generated templates with your modified templates in the catalog, and CSE should be able to use them to deploy clusters that has support for LVM.

In case some extra configuration is required inside the cluster to turn on/manage LVM feature, it can be done by modifying the init/customization scripts viz. https://github.com/vmware/container-service-extension/tree/master/scripts

Regards Aritra Sen

[meltingrobot]

@mbelchin FYI, you can resize the disk without using LVM. You can't join two separate disks, but if you resize the primary disk to a larger size, you can resize the partitions to use the additional space using parted. We have to do that on our cluster currently as the disks in the template given by our provider are way too small.

[mbelchin]

@meltingrobot Thanks for your comment. I already knew about that option and that's exactly what I did just right after creating the cluster. However that's not an option for me. I wouldn't do that on a production cluster. You could have issues resizing disks on a production cluster. LVM is much more flexible and powerful on that sense. I'm using LVM with Kubernetes clusters for several years already and never had an issue.