Closed c0c0n3 closed 1 year ago
Notice there's no clue in the logs about why DirectPV can't pick up the two drives
$ kubectl -n directpv logs deployment/controller
$ kubectl -n directpv logs daemonset/node-server -c node-server
Also, you still get a timeout even if you try explicitly naming the drives
$ kubectl directpv discover --drives=vd{b...c}
Partitioning the drives has no effect either
$ sudo -i
$ parted -a optimal /dev/vdb -- mklabel gpt
$ parted -a optimal /dev/vdb -- mkpart primary 0% 100%
$ parted -a optimal /dev/vdc -- mklabel gpt
$ parted -a optimal /dev/vdc -- mkpart primary 0% 100%
Ditto for formatting with e.g. ext4
.
Possibly all this has to do w/ replacing the original DirectPV x86 images w/ arm64 ones. In fact, Direct PV 4.0.6
can only generate manifests for x86_84/amd64. So to install it I had to
$ kubectl directpv install -o yaml > base.yaml
then replace the images with (what I thought to be) aarch64 equivalent ones and finally kubectl apply
the modified manifest. I saved the modified manifest in base.arm64.yaml
, diff it w/ base.yaml
to see what changed.
re: arm64 install, I opened an issue about it in the DirectPV repo:
So DirectPV does not support ARM64---see https://github.com/minio/directpv/issues/817.
I've deleted the base.arm64.yaml
file, but for the record the only things different from base.yaml
were the images. I replaced
with, respectively
Can't get DirectPV
4.0.6
to work with K8s1.27.2
on NixOS/Aarch64. It looks like DirectPV can't discover available drives and eventually times out.How to reproduce
First off, do a fresh install of our QEMU dev VM for Aarch64. Then create two raw disks
and start the VM with these two additional disks
Finally install DirectPV using our own manifests for Aarch64
Now ask DirectPV to discover any available drives
After a few minutes you should see this error message