[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] Kernel upgrade in domU


  • To: "Fajar A. Nugraha" <list@xxxxxxxxx>
  • From: "Guilherme E.J." <guilherme.e@xxxxxxxxx>
  • Date: Tue, 16 Nov 2010 17:04:52 -0200
  • Cc: xen-users@xxxxxxxxxxxxxxxxxxx
  • Delivery-date: Tue, 16 Nov 2010 11:06:27 -0800
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; b=JPYS/XchvNWbPyx6fvtQtXPeZ4yO06WLN79zRQFX10nh6oOxX6M9nw/gXwwNhQAc6U 6eQWFF69kd651i8fb+nXm0Rh5hFLvZ+vw0KcE2D5TD9hoHyQSwmwRaC0pH/rG4Ng7h8L UqP+xFPDhQji40MIFzRRjgCzL8P+DRQZfdEds=
  • List-id: Xen user discussion <xen-users.lists.xensource.com>

Tks, Fajar!

I'm new on xen, and I wouldn't know pygrub. Really I was making it harder then it should be. ;)

Guilherme



On Tue, Nov 16, 2010 at 3:20 PM, Fajar A. Nugraha <list@xxxxxxxxx> wrote:

On Wed, Nov 17, 2010 at 12:02 AM, Guilherme E.J. <guilherme.e@xxxxxxxxx> wrote:
Hello,

I administrate a lot of xen servers with some linux virtual machines (paravirtualized), and I've some operational problems (costs) to mass upgrade kernel version in domU's.

In a linux server farm (non virtualized servers), the kernel upgrade is trivial, fast and automatized. I simple schedule in my cfengine server, to run "yum update kernel -y && reboot" in all servers.

In a virtualized linux server farm, it's a hard work to upgrade kernel, when needed. First of all, I use cfengine to update kernel-xen package, but after that I've a lot of manual steps that the Operations Department runs:

- Shutdown all virtual machines of xen server X;
- Upgrade kernel and ramdisk of all instances (for example, /xen/hosts/user/i-xxxxx/kernel and /xen/hosts/user/i-xxxxx/ramdisk))
- Upgrade kernel version of dom0
- Reboot dom0

The problem is that I've almost 300 virtual servers, and when we need to upgrade kernel, operations department spend some days to do that.


That's because you're making it harder then it should be.

First of all, if you use pygrub/pv-grub, kernel and ramdisk images reside within domU. So you could ssh (or schedule it with cfengine or whatever) to each domU/instace and run what you usually do on physical machines (changing the kernel package name) : "yum update kernel-xen -y && reboot". No additional steps required to upgrade a domU. That way you can do a rolling upgrade and reboot domU's one by one instead of having to shutdown everything before upgrading.

Second, dom0 should be a minimal installation, with management interface located on private, secure network. In that setup, it should be "safe enough" not having to update the kernel package for dom0 everytime a bug/security fix comes out. So you should be able to schedule dom0 kernel upgrade/reboot (for example) once every six month.

Third, if you really can't affor reboots but can afford monthly subscription, you might want to look at ksplice: http://www.ksplice.com/http://www.ksplice.com/uptrack/supported-kernels 

-- 
Fajar



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.