WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-users

[Xen-users] Questions about fast clock in 64-bit linux hvm

To: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: [Xen-users] Questions about fast clock in 64-bit linux hvm
From: Brendan Wood <bwood@xxxxxxxxxxxxx>
Date: Mon, 21 Jan 2008 19:07:46 -0800
Delivery-date: Mon, 21 Jan 2008 19:08:22 -0800
Envelope-to: www-data@xxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-users-request@lists.xensource.com?subject=help>
List-id: Xen user discussion <xen-users.lists.xensource.com>
List-post: <mailto:xen-users@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
User-agent: Mutt/1.5.17 (2007-11-01)
Hello all,

I've got a couple of questions about an issue I'm seeing in Linux HVM 
guests on 64-bit AMD systems. The clock is running slightly fast, and it 
builds over time at different rates on different guests to become an issue 
in virtual server pods we're testing. I upgraded to Xen 3.2.0 from 3.1.2 
recently but the issue hasn't improved so far. I have a couple of specific 
questions about the problem:

1) I see that there are updates to timer modes for different HVM guest 
OSes. There isn't any documentation anywhere, but I tracked down the 
"timer_mode" xm domain config variable, and found the explanations of the 
different values in xen/include/public/hvm/params.h. It looks like the 
options are delay_for_missed_ticks (default, 0), no_delay_for_missed_ticks 
(1), no_missed_ticks_pending (2) and one_missed_tick_pending (3). I tested 
0 and 1 so far. I saw good results on one domain that hadn't had much skew 
in the first place, but the second one I tested was still a second fast 
soon after I created it. Does anywone have information on what mode works 
best with 64-bit linux?

2) These hosts are already running ntpd, but it doesn't seem to be working. 
Are there particular settings I can use or is there a problem with certain 
versions of ntpd that will prevent it from working correctly in HVM virtual 
OS instances? Using ntp to correct the problem isn't really a fix for the 
issue, but it would be a workaround for the moment while I look for actual 
solutions.

3) In the meantime, if the other options don't work, it looks like I will 
have to run a cronjob using hwclock to reset the system time to the 
"hardware clock," which seems to be the system time of the dom0. Is this 
the case, and if so can I just run ntpd on dom0 and sync the domU times 
periodically using hwclock? This is a gross, herrible hack on my part and 
I'd prefer not to use it, but would it be considered viable in any way?


If anyone has any information or experiences with problems similar to mine, 
I'd appreciate any input. I've been playing with this for some time and 
while it keeps looking like the issue is improving, it's never quite there 
yet. I really want to push Xen for more non-production usage in my company, 
but I just need to iron out these few kinks beforehand. Thanks for any help 
you can give.

Brendan Wood
Responsys, Inc

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

<Prev in Thread] Current Thread [Next in Thread>
  • [Xen-users] Questions about fast clock in 64-bit linux hvm, Brendan Wood <=