Passing a GPU through to a Proxmox container 的修订

albert 修订了这个 Gist 1703420886. 转到此修订

1 file changed, 3 insertions, 1 deletion

tutorial.md

			@@ -78,4 +78,6 @@ lxc.mount.entry = /dev/nvram dev/nvram none bind,optional,create=file
78	78		- Reboot!
79	79		- See if the driver is working
80	80		`nvidia-smi`
81		-	## Now you can install Jellyfin and use your nVidia GPU for transcoding.
	81	+	## Now you can install Jellyfin and use your nVidia GPU for transcoding.
	82	+
	83	+	Video tutorial: https://www.youtube.com/watch?v=-Us8KPOhOCY

albert 修订了这个 Gist 1703420860. 转到此修订

1 file changed, 81 insertions

tutorial.md(文件已创建)

		@@ -0,0 +1,81 @@
1	+	We will pass a GPU through to a proxmox container and utilize it for video transcoding.
2	+
3	+	In PVE Host shell
4	+
5	+	`uname -r`
6	+
7	+	- Install headers
8	+	`apt install pve-headers`
9	+	- Blacklist nouveau
10	+	`nano /etc/modprobe.d/blacklist.conf`
11	+	- Add lines
12	+	`blacklist nouveau`
13	+	- Update Initramfs
14	+	`update-initramfs -u`
15	+	- Reboot!
16	+
17	+	- Install dependencies
18	+	`apt install build-essential`
19	+	- Download driver (https://www.nvidia.com/Download/index.aspx?lang=en-us)
20	+	`wget {YOUR-DRIVER-URL)`
21	+	- Make the driver file executable
22	+	`chmod +x {YOUR-DRIVER-FILE}`
23	+	- Install the driver
24	+	`./{YOUR-DRIVER-FILE}`
25	+	- Make sure the driver loads when restarted
26	+	`nano /etc/modules-load.d/modules.conf`
27	+	- Add lines
28	+	```
29	+	nvidia
30	+	nvidia-modeset
31	+	nvidia_uvm
32	+	```
33	+	- Update Initramfs
34	+	`update-initramfs -u`
35	+	- Create udev rules
36	+	`nano /etc/udev/rules.d/70-nvidia.rules`
37	+	- Add lines
38	+	```
39	+	KERNEL=="nvidia", RUN+="/bin/bash -c '/usr/bin/nvidia-smi -L && /bin/chmod 666 /dev/nvidia*'"
40	+	KERNEL=="nvidia_modeset", RUN+="/bin/bash -c '/usr/bin/nvidia-modprobe -c0 -m && /bin/chmod 666 /dev/nvidia-modeset*'"
41	+	KERNEL=="nvidia_uvm", RUN+="/bin/bash -c '/usr/bin/nvidia-modprobe -c0 -u && /bin/chmod 666 /dev/nvidia-uvm*'"
42	+	```
43	+	- Reboot!
44	+
45	+	- Check if the driver is running
46	+	`nvidia-smi`
47	+	- Edit the container conf you want to passthrough to
48	+	`nano /etc/pve/lxc/{YOUR-LXC-ID}.conf`
49	+	- Note the output of this
50	+	`ls -l /dev/nv*`
51	+	- Add these lines but make sure the path (numbers) are correct
52	+	```
53	+	#Allow cgroup access
54	+	lxc.cgroup2.devices.allow = c 195:0 rw
55	+	lxc.cgroup2.devices.allow = c 195:255 rw
56	+	lxc.cgroup2.devices.allow = c 195:254 rw
57	+	lxc.cgroup2.devices.allow = c 509:0 rw
58	+	lxc.cgroup2.devices.allow = c 509:1 rw
59	+	lxc.cgroup2.devices.allow = c 10:144 rw
60	+
61	+	#Pass through device files
62	+	lxc.mount.entry = /dev/nvidia0 dev/nvidia0 none bind,optional,create=file
63	+	lxc.mount.entry = /dev/nvidiactl dev/nvidiactl none bind,optional,create=file
64	+	lxc.mount.entry = /dev/nvidia-modeset dev/nvidia-modeset none bind,optional,create=file
65	+	lxc.mount.entry = /dev/nvidia-uvm dev/nvidia-uvm none bind,optional,create=file
66	+	lxc.mount.entry = /dev/nvidia-uvm-tools dev/nvidia-uvm-tools none bind,optional,create=file
67	+	lxc.mount.entry = /dev/nvram dev/nvram none bind,optional,create=file
68	+	```
69	+	- Start the container
70	+	- In LXC shell
71	+	`apt update && apt upgrade -y`
72	+	- Download driver (https://www.nvidia.com/Download/index.aspx?lang=en-us)
73	+	`wget {YOUR-DRIVER-URL)`
74	+	- Make the driver file executable
75	+	`chmod +x {YOUR-DRIVER-FILE}`
76	+	- Install the driver
77	+	`./{YOUR-DRIVER-FILE} --no-kernel-module`
78	+	- Reboot!
79	+	- See if the driver is working
80	+	`nvidia-smi`
81	+	## Now you can install Jellyfin and use your nVidia GPU for transcoding.