Resizing of LVM after Host-Reboot not working #647
Comments
|
Hi! Thanks for reporting this issue. We are currently investigating quite similar issues. Our best guess is that there is a race between the container lvmtools (where we completely disable udev, so However, we have not been able to trace the exact reason why a resize or taking a snapshot can cause udevd to remove the symlink from the list. |
|
Could you give the following configuration a try? ---
apiVersion: piraeus.io/v1
kind: LinstorSatelliteConfiguration
metadata:
name: udev
spec:
podTemplate:
spec:
containers:
- name: linstor-satellite
volumeMounts:
- name: lvmconfig
mountPath: /etc/lvm/lvm.conf
subPath: lvm.conf
readOnly: true
volumes:
- name: lvmconfig
configMap:
name: lvmconfig
---
apiVersion: v1
kind: ConfigMap
metadata:
name: lvmconfig
data:
lvm.conf: |
activation {
udev_sync=1
monitoring=0
udev_rules=1
}
devices {
global_filter="r|^/dev/drbd|"
obtain_device_list_from_udev=1
}We already pass through the udev socket so enabling the lvmtools to wait for udev should not be an issue. That way there should be no race between container lvm tools and udevd |
|
This configuration is working! But in the errlog I get following: |
|
Could that be remains from a previous attempt? I.e. can you try a new resize now? Do these errors still happen? |
|
No this was a new attempt. And a new resize cause the same Error |
|
Have you rebooted the nodes in the meantime? I guess an sos-report would be good. |
|
Now I rebooted the nodes and the same error |
|
Ok, so one small fix is also settings Haven't been able to find the source of your specific issues, but this fixes lvm commands hanging on lvcreate, etc.. |
|
This small fix help! |
|
I have the same problem. |
Because making LVM work consistently when running in a container is hard. Even harder if the solution should support Ubuntu, Debian, RHEL, Talos Linux, and whatever else might be running out there. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi,
I am using Linstor with Piraeus-Operator (2.5.0) on my Kubernetes cluster. The OS is Ubuntu 22.04.4 with lvm installed and Kubernetes is the distro from Rancher, v1.27.12+rke2r1
My problem is nearly the same problem as described in this feature request LINBIT/linstor-server#326 .
As long as I have all nodes running I can create and resize disks as often as I want. But as soon as I restart the nodes I have exactly the same problem as described above. But what I see is the following difference.
After some research I see that after reboot udev see the lvms and create the symlinks to the dm devices. And there is the Problem. After this action the lvresize in the container can resize this volumes but lost the symlinks.
Could it be that there is a general problem, or have I done something fundamentally wrong?
Even with manual tests with other distros (Suse Enterpris Linux Micro 5.4 --> generated with Rancher Elementel) I encounter the same problem.
If more details are needed from me, please let me know.
The text was updated successfully, but these errors were encountered: