This message was deleted.
# harvestera
adamant-kite-43734
12/03/2024, 2:26 PMThis message was deleted.
g
glamorous-fireman-38699
12/04/2024, 7:09 AMThis must be a BUG and I created and issue on Github - hope it will be solved or else I can’t use Harvester 😞
g
great-bear-19718
12/07/2024, 12:01 AMif you want devices one a specific node then please also specify host details in scheduling
great-bear-19718
12/07/2024, 12:02 AMwe added logic to reconcile what is actually added when the vm is running
g
glamorous-fireman-38699
12/07/2024, 12:04 AM@great-bear-19718 I have two single nodes, I did at some point have the actual node selected, but I still think it does not work with a specific PCI device order.
g
great-bear-19718
12/08/2024, 8:10 AMi am not sure how pci device order would matter if you have two devices both will be passed to the VM
great-bear-19718
12/08/2024, 8:10 AMissue is how the allocation is performed by the kubelet if you have multiple devices of same kind then it may not be the same as passed as kubelet assigns one based on resource name
great-bear-19718
12/08/2024, 8:10 AMthere is a issue pending which should fix this behaviour
g
glamorous-fireman-38699
12/08/2024, 10:22 AMPCI device order does matter in my case. I am running a virtual firewall and it picks management device and the rest of the NICs in next order, so it has to be the same order or else patching up server is a pain. If it would just follow the array list on the YAML all will be good, but great to hear we might have a fix soon 🙂 - will it be there in next release or is there place I can watch/follow progress?