My idea is to run virtual machines that if hardware fails moves over to some other physical box in case of a hardware failure. Even if you have fast boxes with lots of memory, there is a limit to how many virtual machines you can run on one box, and still get decent performance. So chances are that I will nned more physical boxes. At the very least there should be a plan on how to add them, and it should be possible to expand the system without taking it down.<br>
<br>How long in the future? As soon as possible, or when I'm confident enough to use this in a production environment.. By then I probably move to CentOS RHEL, but I'm used to Fedora so I'll start testing<br>
on that.<br><br>As you ask about the future, are there any new things on the horizon that I should wait for? <br><br><br><br><div class="gmail_quote">2008/5/7 Todd Denniston <<a href="mailto:Todd.Denniston@ssa.crane.navy.mil">Todd.Denniston@ssa.crane.navy.mil</a>>:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Uno Engborg wrote, On 05/07/2008 03:33 AM:<div class="Ih2E3d"><br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
The problem is that it can't be expanded to more than 2 boxes in the future.<br>
<br>
</blockquote>
<br></div>
what do you mean by more than 2 boxes[1], and how far in the future?<br>
<br>
<a href="http://www.linbit.com/en/products-services/drbd/drbd-plus/" target="_blank">http://www.linbit.com/en/products-services/drbd/drbd-plus/</a><br>
I believe they will probably roll this functionality into the openly released version later, but even so they have described on the mailing list that it's setup is very difficult, i.e., you'll want to pay them to set it up.<br>
<br>
[1] with the current setup you can have several nodes in one drbd cluster, but _each_ drbd _resource_ is only mirrored between two nodes.<div><div></div><div class="Wj3C7c"><br>
<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
<br>
Thanks!<br>
/uno<br>
<br>
<br>
2008/5/5, Gijs <<a href="mailto:info@boer-software-en-webservices.nl" target="_blank">info@boer-software-en-webservices.nl</a>>:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
You might want to take a look at <a href="http://www.drbd.org/" target="_blank">http://www.drbd.org/</a>.<br>
It basically does what you suggest. It creates a software raid 1 partition<br>
across a local and networked partition.<br>
Perfect for real-time syncronisation between disks across a network, while<br>
not having to resort to expensive stuff like SANs and the like.<br>
<br>
Regards,<br>
<br>
Gijs<br>
<br>
Uno Engborg wrote:<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
Hi,<br>
I'm experimenting with a HA setup on two physical boxes. The idea is to<br>
create virtual Xen machines that automagically do live migration from one<br>
physical box to the other in case of e.g. a hardware failure on one of the<br>
physial boxes.<br>
<br>
My idea is to create a software raid of networked block devices and real<br>
disks. I.e:<br>
<br>
Box A:<br>
Software raid consisting of:<br>
Physical disk in box A<br>
Nework block device exported from box B<br>
<br>
Box B:<br>
Software raid consisting of:<br>
Physical disk in box B<br>
Network block device exported from box A<br>
<br>
<br>
On top of this I plan t use GFS2 and a cluster that handles the<br>
migration of the virtual machines.<br>
<br>
Any comments on this, would this work, or would it result in deadlock?<br>
What about quorum disk? I suppose<br>
I need one? Most setup of these kind of things usually have their<br>
storage on a separate<br>
SAN, but as this is mostly for experimentation and testing purposes, I<br>
would hope that I could do without that for now.<br>
<br>
What would be the best way to export the network block devices? I'm<br>
thinking iSCSI or GNDB?<br>
<br>
If this doesn't look like a good idea, are there any other way to do<br>
this.<br>
<br>
Regards<br>
Uno Engborg<br>
<br>
<br>
</blockquote>
--<br>
fedora-list mailing list<br>
<a href="mailto:fedora-list@redhat.com" target="_blank">fedora-list@redhat.com</a><br>
To unsubscribe: <a href="https://www.redhat.com/mailman/listinfo/fedora-list" target="_blank">https://www.redhat.com/mailman/listinfo/fedora-list</a><br>
<br>
</blockquote>
<br>
</blockquote>
<br>
<br></div></div><font color="#888888">
-- <br>
Todd Denniston<br>
Crane Division, Naval Surface Warfare Center (NSWC Crane)<br>
Harnessing the Power of Technology for the Warfighter</font><div><div></div><div class="Wj3C7c"><br>
<br>
-- <br>
fedora-list mailing list<br>
<a href="mailto:fedora-list@redhat.com" target="_blank">fedora-list@redhat.com</a><br>
To unsubscribe: <a href="https://www.redhat.com/mailman/listinfo/fedora-list" target="_blank">https://www.redhat.com/mailman/listinfo/fedora-list</a><br>
</div></div></blockquote></div><br>