现在，通过使用pv-grub--GRUB引导程序在Xen mini-os内核中的移植，可以在你的Linode实例下启动自定义内核，甚至是替代的操作系统。 基本上，你的Linode启动到pv-grub内核，寻找你的menu.lst和相关内核文件，然后执行你的内核。
虽然这个功能听起来很整洁，但它绝对是为你的调整者准备的。 除了支持pv-grub本身，我们不会提供任何对自定义配置或替代操作系统的支持。 我们仍然建议你运行我们的内核，除非你有一些非常特殊的需求，或者只是想乱来。
有一篇wiki文章可以让你开始，甚至还有另一篇文章概述了在Linode上释放BSD的情况。
评论 (9)
[…] Posted on December 24, 2008 – 8:19am I reckon the logo on Linode Wiki needs a change. Via Linode’s latest blog post, it is now possible to roll your own operating system with Linode’s pv-grub support. […]
[…] Linode Blog » Custom kernels with pv-grubThis would be enough to run selinux or app-armor on linode now. […]
How stable is that? Is it stable enough for a production system?
How stable is pv_grub? pv_grub boots your own kernel, and then it’s done, so it has little to do with stability. I think the question is: how stable the custom kernel that you’ve provided is.
Well, the reason I’m asking is because I have experienced stability problems with vmware’s CONFIG_VMI kernel option in an otherwise very stable kernel.
In case anyone else is thinking of going down the same path I tried, the CentOS 5.2 kernel-xen package wasn’t built with all of the options specified on the linked wiki page, so it won’t work with pv-grub without a custom compile (unless I did something wrong).
That’s great news. I’m going to play around with this a bit.
I suspect I won’t be able to get it to do what I ultimately want: boot an OpenSolaris (2008.11) instance. I’ve been able to get OpenSolaris running well on XenServer, but I can’t use pv_grub. The root filesystem is ZFS, so the script can’t reach in and pull out the menu.lst file unless it knows how to deal with that FS.
In my experience, I have to use the PV-args and PV-kernel parameters in XenServer to make it work (see my website for details). If there’s a way that you can expose those configuration items, that would be awesome.
I recently switched to Linode because of this. I’ve been running Gentoo with my own custom kernel for a few days now and it’s working great. This feature is awesome.
I’m in the process of switching from slicehost because of this.