--- Matthijs ter Woord <matthijsterwoord@gmail.com> wrote:> Is it possible to eliminate the possibility of a Xen host failure due to > hardware failure (ie. failover clustering)?I''m no Xen expert but from what I can tell you can''t to high-availability with the Xen kernel itself. However, if you think about it, there is very little difference between domU virtual machines and real machines. You can easily do high-availability on domU guests between two physical Xen hosts. Run all of your services on domUs and use something like UltraMonkey, regular LVS, Heartbeat, OpenSSI (assuming it works with the domU kernel patches), or some other HA solution giving you failover clustering. The difference between this and a real cluster is only that Xen is running below the virtual machines; very little is different. They only add a thin layer of separation. So for example if your poison of choice is Heartbeat and you want to make Apache highly available, you do this: Apache1 Apache2 | | +---Heartbeat---+ | | DomU1 DomU2 virtual virtual machine machine | | Xen1 Xen2 kernel kernel | | Node1 Node2 hardware hardware A real cluster would look like this: Apache1 Apache2 | | +---Heartbeat---+ | | Node1 Node2 hardware hardware Very little difference. Hope that helps! CD You have to face a Holy God on Judgment Day. He sees lust as adultery (Matt. 5:28) and hatred as murder (1 John 3:15). Will you be guilty? Jesus took your punishment on the cross, and rose again defeating death, to save you from Hell. Repent (Luke 13:5) and trust in Him today. NeedGod.com _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Yep, Chris''s comments are a good guide - just pretend the Xen virtual machines are real machines. Just to note: there are some people (in Cambridge and elsewhere) working on things like running two *virtual* machines on differenct hosts exactly in-step (at the instruction level) to guard against higher-level failures without needing application-level fail-over software. There''s no code available yet though. Cheers, Mark On Sunday 14 August 2005 23:05, Chris de Vidal wrote:> --- Matthijs ter Woord <matthijsterwoord@gmail.com> wrote: > > Is it possible to eliminate the possibility of a Xen host failure due to > > hardware failure (ie. failover clustering)? > > I''m no Xen expert but from what I can tell you can''t to high-availability > with the Xen kernel itself. > > However, if you think about it, there is very little difference between > domU virtual machines and real machines. You can easily do > high-availability on domU guests between two physical Xen hosts. Run all > of your services on domUs and use something like UltraMonkey, regular LVS, > Heartbeat, OpenSSI (assuming it works with the domU kernel patches), or > some other HA solution giving you failover clustering. > > The difference between this and a real cluster is only that Xen is running > below the virtual machines; very little is different. They only add a thin > layer of separation. > > > So for example if your poison of choice is Heartbeat and you want to make > Apache highly available, you do this: > Apache1 Apache2 > > +---Heartbeat---+ > > DomU1 DomU2 > virtual virtual > machine machine > > Xen1 Xen2 > kernel kernel > > Node1 Node2 > hardware hardware > > A real cluster would look like this: > Apache1 Apache2 > > +---Heartbeat---+ > > Node1 Node2 > hardware hardware > > > Very little difference. > > Hope that helps! > > CD > > You have to face a Holy God on Judgment Day. He sees lust as adultery > (Matt. 5:28) and hatred as murder (1 John 3:15). Will you be guilty? > > Jesus took your punishment on the cross, and rose again defeating death, to > save you from Hell. Repent (Luke 13:5) and trust in Him today. > > NeedGod.com > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Is it possible to eliminate the possibility of a Xen host failure due to hardware failure (ie. failover clustering)? Greets _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> Isn''t there a way to set up a drbd partition which contians the vm harddrive > images, setting files etc, and somehow detect a failure of node 1 so node 2 > can start booting?Yes, but that''s not really failover, eh? :) John -- John Madden UNIX Systems Engineer Ivy Tech Community College of Indiana jmadden@ivytech.edu _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Isn''t there a way to set up a drbd partition which contians the vm harddrive images, setting files etc, and somehow detect a failure of node 1 so node 2 can start booting? Greets, Matthijs ter Woord ----- Original Message ----- From: "Chris de Vidal" <cdevidal@yahoo.com> To: "Matthijs ter Woord" <matthijsterwoord@gmail.com>; <xen-users@lists.xensource.com> Sent: Sunday, August 14, 2005 3:01 PM Subject: Re: [Xen-users] Failover> --- Matthijs ter Woord <matthijsterwoord@gmail.com> wrote: > > Is it possible to eliminate the possibility of a Xen host failure due to > > hardware failure (ie. failover clustering)? > > I''m no Xen expert but from what I can tell you can''t to high-availabilitywith> the Xen kernel itself. > > However, if you think about it, there is very little difference betweendomU> virtual machines and real machines. You can easily do high-availabilityon> domU guests between two physical Xen hosts. Run all of your services ondomUs> and use something like UltraMonkey, regular LVS, Heartbeat, OpenSSI(assuming> it works with the domU kernel patches), or some other HA solution givingyou> failover clustering. > > The difference between this and a real cluster is only that Xen is running > below the virtual machines; very little is different. They only add athin> layer of separation. > > > So for example if your poison of choice is Heartbeat and you want to make > Apache highly available, you do this: > Apache1 Apache2 > | | > +---Heartbeat---+ > | | > DomU1 DomU2 > virtual virtual > machine machine > | | > Xen1 Xen2 > kernel kernel > | | > Node1 Node2 > hardware hardware > > A real cluster would look like this: > Apache1 Apache2 > | | > +---Heartbeat---+ > | | > Node1 Node2 > hardware hardware > > > Very little difference. > > Hope that helps! > > CD > > You have to face a Holy God on Judgment Day. He sees lust as adultery(Matt. 5:28) and hatred as murder (1 John 3:15). Will you be guilty?> > Jesus took your punishment on the cross, and rose again defeating death,to save you from Hell. Repent (Luke 13:5) and trust in Him today.> > NeedGod.com_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Looks much like it. All the info i can find about Xen is all about high availability for maintenance downtime etc. But that''s one of the least important things to me about high availability. Please note that i don''t have experimented with Xen yet. Lets assume (don''t pin me to it, as i''m just assuming these numbers) that an average xen virtual machine takes 1 minute to boot up. Then it would only cost me 1 minute downtime when a server crashes. But as i read the info i''ve seen so far about Xen, there''s even not mentioned a hardware failure. Greets, Matthijs ter Woord ----- Original Message ----- From: "John Madden" <jmadden@ivytech.edu> To: "Matthijs ter Woord" <matthijsterwoord@gmail.com> Cc: <xen-users@lists.xensource.com>; "Matthijs ter Woord" <matthijsterwoord@gmail.com> Sent: Monday, August 15, 2005 5:53 AM Subject: Re: [Xen-users] Failover> > Isn''t there a way to set up a drbd partition which contians the vmharddrive> > images, setting files etc, and somehow detect a failure of node 1 sonode 2> > can start booting? > > Yes, but that''s not really failover, eh? :) > > John > > > > > > -- > John Madden > UNIX Systems Engineer > Ivy Tech Community College of Indiana > jmadden@ivytech.edu >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users