any plans to support nvme tcp on longhorn and nodes?

I noticed something unexpected with the v1.1.2 ISO installer. I installed v1.1.2 on a different disk on the same machine where v1.1.1 had previously been installed, and the previous disk was silently, unexpectedly re-used by the new install as a persistence volume.

We set quotas for users so that they can only use certain amounts of CPU/memory/storage. The quotas are correctly enforced but the users don't see when they are over the quota. The VM is simply not started. Even admins don't have a visible indicator. The only way to find out is to check the kubernetes events for scheduler errors. How is this supposed to work? How is a normal user supposed to identify such an issue?

Anyone using Ryzen/Epyc based CPU ? Seems that nested virtualization does not work at all in 1.1.2 https://github.com/harvester/harvester/issues/3900

Hi, is it ok to share a technical 'solution' i've done for quick feedback if i've set myself up for failure (Booting Harvester off a USB/SD-Card in HP Dl servers, but run harvester off an nvme/sata drives). I got this to work, but i'm debating if i've set myself up for a crash at he next upgrade 😄

I have an older DL360 Gen8 which has a notorious raid card forbidding HBA/IT mode, or rather you cannot boot if enabled. The default install of Harvester, even the installation disk has heavy I/O. I started to do chroot/grub, but honestly had not messed with image/immutable OS boots before and could not get the chroot properly done. So.. I installed Harvester once to the USB drive itself. Then installed to just the nvme drive, by itself. Which is not bootable by the server. After booting up(Coming into the USB installation), I changed the labels of COS_STATE and COST_STATE_PERSISTENT, and HARV_LH_DEFAULT on the USB drive, so that the nvme partitions would be found instead, containing the correct labels. This now boots properly, and seeming 0 I/O on the USB stick. Curious if this is clever or setting myself up for a massive failure during an upgrade -- thinking harvester uses some of the cfg files from the initial to know where to expect things when upgrading?

How can add NIC ( which is basically cable from the internet, so its untagged ) to the Harvester? I have only mngt (eth18) NIC available for adding cluster network, see diagram for more info https://kroki.io/mermaid/svg/eNpLy8kvT85ILCpRCHHhUlBQSC3JMLQIyi8tSS1S0NXTtavJzUvP[…]SYHQsVtEQStIzlSs1LQdYPsQhhAISPbgKSKIYRcM_CDYGLoBmDKg4ziAsA4fiJ8A Basically, I would expect that I can add a network with a public IP range. Thanks

Hi, I am hoping someone can provide some insights - I am running into a problem installing harvester in my environment - the installer completes and the system reboots, however the Status on the console is stuck on "Setting up Harvester". No VIP gets assigned even though I designated a static VIP during install. Below is the kubectl get service -A output - are there particular logs i should be looking at or is there a document that describes the boot process so I can look into where it's getting stuck?

image.png

need some help... both nodes are schedulable but it still says ''replica scheduling failed''

storage class config:

How can I do a volume with 2 separate servers? like I have 2 servers with 300gb, how do I cluster the storage to make 1 volume?

hello, I got an occasional failure when I reboot VM assigned with a GPU device: 0/4 nodes are available: 4 Insufficient nvidia.com/TU102_GEFORCE_RTX_2080_TI_REV__A. preemption: 0/4 nodes are available: 4 No preemption victims found for incoming pod. I'm sure that the GPU card is sufficient because everything is ok before reboot VM. The failure disappear after rebooting this harvester node which version is 1.1.1.

Did anyone find a way to implement backup&restore using a Veeam backup repository? Or is it possible to install Kasten K10 in the main cluster?

Hi all! Can harvester’s disk share with each other？

Hi Friends. i have some built-in rancher/harvester deployments/statefulsets that keep deploying on a node i dont want them on.

StatefulSet: prometheus-rancher-monitoring-prometheus

if i go into rancher, adn set the node scheduling, it just resets it to nothing, how can i apply thing change and it stick?

hi harvester team, right now the gpu device would be pass to vm via pci passthrough, is it possible to run time-sliced gpu to vm instead? so that we would share the same gpu for multiple vms

hi everyone, I plan to build my own infrastructure and want to experiment with a harvester. I am new to this stuff so my question can be a bit basic. For the hardware, is there any minimum node to be applied? and each node needs how many cores minimum? I checked the documentation and it said 16-cores and above? Also, for the memory, do we need to separate data and OS to different SSD/NVME, or do partition in a single memory is ok-ish?

Is there somewhere I can read about Longhorn's methodologies for replication and where volumes reside, etc, on Harvester?

hey guys, I have few more questions: 1. Can Harvester use two different types of storage, such as using SSD for the Harvester OS and HDD for data instances, or does it require using one type only (all SSD or all HDD)? 2. If we install more than one storage disk, does the system need to operate in RAID or can it be used in JBOD (Just a Bunch of Disks) mode without RAID?

Hi, I am trying to run Harvester ISO on a baremetal machine. But one of the pod from deployment “harvester-cluster-repo” is failing with the below error. Seems like the used docker image is not there on docker.io .

Is there a way to re-add a host after it has been deleted without reinstalling?

having some issues with a rancher managed k8s cluster running on a harvester cluster – one of the nodes is stuck with InfrastructureReady = False, even though the node shows up correctly:

- lastTransitionTime: '2023-05-24T14:30:22Z'
      message: >-
        creating server [fleet-default/test-main-1-1e68de03-7hcf6] of kind
        (HarvesterMachine) for machine test-main-1-7d6f6d4f4b-nhhpt in
        infrastructure provider
      status: 'False'
      type: InfrastructureReady

Hello, In an RKE cluster (generated by Harvester 1.1.2) we have now stuck master nodes in removing state. (This happened since the VMs were removed) If we run kubectl against the still alive masters they work, In “get nodes” the removed masters are not shown. However the cluster API is not OK and we see this state (screenshot) How can we “force” remove them ?

I have an issue on upgrade from v1.1.0 -> 1.1.2 where the mcc-rancher-monitoring is modified and holding up the process. I for the life of me cannot figure out how to fix the mcc-rancher-monitoring. I've done a few rollbacks on the helm chart but that didn't work. I also manually updated the statefulset for prometheus to correct the incorrect values. I've been through all of the issues that are in any way related to this. They're helpful in that I understand what's wrong but at the same time I'm haven't been able to get the bundle fixed. Any ideas?

NAMESPACE     NAME                                          BUNDLEDEPLOYMENTS-READY   STATUS
fleet-local   fleet-agent-local                             1/1                       
fleet-local   local-managed-system-agent                    1/1                       
fleet-local   mcc-harvester                                 1/1                       
fleet-local   mcc-harvester-crd                             1/1                       
fleet-local   mcc-local-managed-system-upgrade-controller   1/1                       
fleet-local   mcc-rancher-logging                           1/1                       
fleet-local   mcc-rancher-logging-crd                       1/1                       
fleet-local   mcc-rancher-monitoring                        0/1                       Modified(1) [Cluster fleet-local/local]; <http://prometheus.monitoring.coreos.com|prometheus.monitoring.coreos.com> cattle-monitoring-system/rancher-monitoring-prometheus modified {"spec":{"resources":{"limits":{"memory":"2500Mi"},"requests":{"memory":"1750Mi"}}}}
fleet-local   mcc-rancher-monitoring-crd                    1/1

How can I remove these masters ? (they no longer exist)

managed to fix it by restarting the Rancher pods

hi team, anyone has successfully configured harbor (container registry) inside a vm? I tried one with network mode (bridge), it only work during the second time of pulling image( the first time is always stucked). But the other vm with masquerade are not working at all.