cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From jerry jiang. 蒋维 <jerry.ji...@tyxtech.com>
Subject Re: overcommit radio setting per-pod basis and restoring VMs if host failed
Date Thu, 27 Dec 2012 13:29:56 GMT
Hi Nithin

Thanks.
My comments inline.


----- Original Message -----
From: "Nitin Mehta" <Nitin.Mehta@citrix.com>
To: cloudstack-users@incubator.apache.org
Sent: Thursday, December 27, 2012 8:53:48 PM
Subject: Re: overcommit radio setting per-pod basis and restoring VMs if host failed

Answers inline

On 27-Dec-2012, at 5:58 PM, jerry jiang. 蒋维 wrote:

> hello,
> 
> I have two questions regarding Cloudstack function,
> 
> 1) How can I set different overcommit radio for different pod (or cluster) in one zone.

Currently this is not possible. But, there is a plan for this feature in the upcoming release
for setting it cluster based. If you don't mind, can you please share your use case for setting
it different for different pods ?

[jerry] I want this feature because I want to deploy two pods for different department in
my datacenter. One is for testing my Apps to verify its functions, I would like this pod to
support overcommit radio(120%) because function-testing won't consume much CPU resource. and
another one is for testing my Apps to verify performance, I would like this pod not to support
undercommit radio(80%) becasue it will guarentee the Apps to get resource as needed.


> 
> 2) When I use CS to manage hosts, if one of host failed, can I restore the VMs in the
host via providing another healthy host?
> 

While creating vm's you can use service offering which has HA (highly available) as true.
If the host fails CS will automatically take care of migrating it to a healthy host within
the cluster.
Let's say you didn't use that service offering while deploying the vm then if the host fails,
these vms would go into stopped state. You can try starting these vms and they should come
up on a healthy host within the cluster. I am assuming you are using shared storage here and
would have enough capacity in the cluster.

[jerry] that means I can create VMs by API directly even if I didn't create service offering,
right? please comfirm.


> Thanks
> 
> Jerry Jiang


Mime
View raw message