lyulyul / shine-cluster

Simple High performance Infrastructure for Neural network Experiments
GNU General Public License v3.0
14 stars 8 forks source link

Experiment remote virtual machines #183

Open gqqnbig opened 1 year ago

gqqnbig commented 1 year ago

We have virtual machines available on aha (https://github.com/luoyuqi-lab/shine-cluster/wiki/%E8%99%9A%E6%8B%9F%E6%9C%BA%EF%BC%88%E5%AE%9E%E9%AA%8C%E6%80%A7%EF%BC%89). But if my memory servers, it was hard to use. I had to use x11 for VirtualBox windows and they had display glitches.

https://www.virtualbox.org/manual/ch07.html seems to indicate VirtualBox is able to run remotely. In that case, a user can run VirtualBox client on his workstation and the client connects to the Shine cluster. I bet the user experience will be smoother.

Moreover, VirtualBox is freeware while VMWare is paid. I don't believe many users actually paid for it.

When testing new software or experimenting runtime (except for GPU tasks), it's easier, safer, and better to do it in a virtual machine. In light of the 16GB memory of the stock workstations, providing VM capability certainly gives Shine cluster users a head start over other researchers.

gqqnbig commented 1 year ago

Find OS images

https://www.ventoy.net/en/distro_iso/windows.html

https://github.com/pbatard/Fido

https://files.rg-adguard.net/version/f0bd8307-d897-ef77-dbd6-216fefbe94c5

dancingBone79 commented 1 year ago
  1. download https://sourceforge.net/projects/vcxsrv/, and install it.
  2. 打开patty,导入aha,并选择 SSH 下 X11 ,勾选X11 forwarding
  3. patty 的终端中 Linux命令行输入export LIBGL_ALWAYS_INDIRECT=1,再输入 gedit, 最后输入virtualbox 启动软件
  4. According to : https://www.virtualbox.org/manual/ch07.html, set the virtualbox. 4.1 Installing Oracle VM VirtualBox and Extension Packs, 在客户端和服务端都需要安装好,且需要版本完全一致 Oracle VM VirtualBox: The ubuntu(.ios) need to be downloaded as well, and set as the vm source. Extension Packs Link:https://www.virtualbox.org/wiki/Download_Old_Builds_6_1 install command: sudo VBoxManage extpack install Oracle_VM_VirtualBox_Extension_Pack-6.1.16.vbox-extpack ( If mixmatched, pls uninstall the old one: sudo VBoxManage extpack cleanup sudo VBoxManage uninstall "Oracle VM VirtualBox Extension Pack" )

    Remember to restart both of the software (server and client), and wait for some minutes.

4.2 Set host IP address: 3389(change into others, eg:53389) Display --> remote Display --> server port[...], method[NULL], timeout[5000]

  1. Install Remmina 需要在你用来远程访问其它电脑的的主机上安装 Remmina Command link: (https://remmina.org/how-to-install-remmina/#ubuntu)

  2. Click the green button on the Oracle VM VirtualBox to start the connection

gqqnbig commented 1 year ago

putty_TwyzgqArow

I paused my vm and they are not running, but there are still two process wasting our precious memory.

一个虚拟机实例的连续运行时间受制于集群简介#运行时间限制 (ref)

Your guide asks us users not to run vm overtime. Although I paused my vm promptly, this VBoxHeadless continues running for 51 hours. Will it be killed? Will killing the processes cause my data loss? Did I pause my vm in a wrong way?

See also https://forums.virtualbox.org/viewtopic.php?t=104007

dancingBone79 commented 1 year ago

putty_TwyzgqArow

I paused my vm and they are not running, but there are still two process wasting our precious memory.

一个虚拟机实例的连续运行时间受制于集群简介#运行时间限制 (ref)

Your guide asks us users not to run vm overtime. Although I paused my vm promptly, this VBoxHeadless continues running for 51 hours. Will it be killed? Will killing the processes cause my data loss? Did I pause my vm in a wrong way?

See also https://forums.virtualbox.org/viewtopic.php?t=104007

Maybe it's a bug from VBoxManager, (ref) and (ref). And there's also a sound said this problem was introduced between 6.1.24 and 6.1.26. and should be fixed in 6.1.27r147422 and newer.