I've finally gotten my first node up and running. This is on a RHEL6.3 system, I have a script which, once I'm done, I'll post on my blog.
In the meantime, I want to add a 2nd node. Once I've created it and added it, how do I change the existing VMs to be a DRDB setup so that I can migrate them from one system to another?
On Thu, Sep 27, 2012 at 05:17:59PM -0400, Jonathan Bayer wrote:
> I've finally gotten my first node up and running. This is on a
> RHEL6.3 system, I have a script which, once I'm done, I'll post on
> my blog.
> In the meantime, I want to add a 2nd node. Once I've created it and
> added it, how do I change the existing VMs to be a DRDB setup so
> that I can migrate them from one system to another?
<linuxguruga...@gmail.com> wrote:
> I've finally gotten my first node up and running. This is on a RHEL6.3
> system, I have a script which, once I'm done, I'll post on my blog.
> In the meantime, I want to add a 2nd node. Once I've created it and added
> it, how do I change the existing VMs to be a DRDB setup so that I can
> migrate them from one system to another?
for every instance you that you want to convert. There will be
downtime for each instance while data is copied to the second node.
Somewhat riskier way is to run 'gnt-instance modify -t drbd -n
$second_node --no-wait-for-sync $instance' instead. Then you can start
instance earlier before data is fully copied, but you may have worse
performance due to heavy I/O and potentially higher chance of losing
instance data if something goes wrong.
On Thu, Sep 27, 2012 at 06:06:50PM -0400, Kirill Shevchuk wrote:
> On Thu, Sep 27, 2012 at 5:17 PM, Jonathan Bayer
> <linuxguruga...@gmail.com> wrote:
> > I've finally gotten my first node up and running. This is on a RHEL6.3
> > system, I have a script which, once I'm done, I'll post on my blog.
> > In the meantime, I want to add a 2nd node. Once I've created it and added
> > it, how do I change the existing VMs to be a DRDB setup so that I can
> > migrate them from one system to another?
> for every instance you that you want to convert. There will be
> downtime for each instance while data is copied to the second node.
> Somewhat riskier way is to run 'gnt-instance modify -t drbd -n
> $second_node --no-wait-for-sync $instance' instead. Then you can start
> instance earlier before data is fully copied, but you may have worse
> performance due to heavy I/O and potentially higher chance of losing
> instance data if something goes wrong.
Kirill, that's a very nice explanation. Would you take the time to send
a patch against the docs I quoted in the other reply or against the man
pages to make more clear what are the trade-offs?
