Community
 
 
 

CloudPlatform 3.x

343フォロワー
 
Avatar
Pankaj Paliwal

Compute Issues after Installation and adding hosts and compute on CloudStack3

Avatar

Compute Issues after Installation and adding hosts and compute on CloudStack3

Hi All,

I have following setup:

CloudStack 3 latest till date of writing on Ubuntu 10.04 LTS 64-Bit

Total 2 Nodes

1 is the management node and compute node as well (installed both management, database and agent(for compute)
1 is compute node only

1 Zone, 1 Pod, 1 Cluster and 2 Hosts in cluster.

Using KVM on both machines

IP address scheme used: 192.168.1.0/24

All Hosts added and I can access the management web portal as well and it is showing 22GHz CPU and 14GB Ram in Infrastructure as well with both hosts up and running.

I can ping both of them and mount the nfs on both without any issues.

When I try to create an instance with Medium , Small or any other that has 2 Cores assigned with 2GHz each and 2GB Ram it works fine if I try to assign any thing more than that it gives error on creating the Instance (VM)

Management and Compute node: Intel Core i5 3.10GHz , 8GB Ram, 300GB HDD, Intel Mobo

Compute Node: Intel Core i5 2.4GHz, 6GB Ram Memory, 640GB HDD, Intel Mobo

Please advise what shall I do now to create one big instance having 6 or 8 cores with atleast 15-16GHz performance and 8-10GB Ram memory.

Edited by: engrusmanmalik on Oct 14, 2012 7:24 AM

Attached Files


Usman Malik MEMBERS
3件のコメント
0

サインインしてコメントを残してください。

 
 

Previous 3件のコメント

Avatar
Pankaj Paliwal
Avatar

Compute Issues after Installation and adding hosts and compute on CloudStack3

> Please advise what shall I do now to create one big instance having 6 or 8 cores with atleast 15-16GHz performance and 8-10GB Ram memory.
Your host needs enough resources to boot the VM with the configuration in the Compute Offering. CloudStack does not support memory over-provisioning, so if your host only has 8 GB RAM, you won't be able to deploy a VM with a Compute Offering with 8 GB RAM or more due to overhead (the host itself uses some RAM) and other VMs using RAM (even VMs that aren't running will take up capacity, which is controlled by capacity.skipcounting.hours in Global settings). CloudStack does support CPU over-provisioning, but since the Core i5 is only a quad-core CPU, you won't be able to deploy a VM using a Compute Offering with more than four cores.

When deploying a Compute Offering that is too big, the management-server.log will have an entry like:
> 2012-10-15 14:24:50,043 DEBUG \[cloud.deploy.FirstFitPlanner] (Job-Executor-1:job-189) No clusters found having a host with enough capacity, returning.

The same error may appear for other problems, but you can look through the other log entries for the job and see the root cause.

Best regards,

{color:#555555}Kirk Kosinski{color} !http://www.linkedin.com/favicon.ico!
{color:#999999}MCITP: EA / VA / EDA7, VCP 4 / 5, CCA{color}


Kirk Kosinski CITRIX EMPLOYEES
コメントアクション パーマリンク
Avatar
Pankaj Paliwal
Avatar

Hi Kirk,

Thanks for you reply.

I understand that CloudStack does not support CPU and Memory over-provisioning.

This means that if it is showing "22 GHz" in infrastructure overview layout i can not create a big compute VM instance and would have to stick to the original host restrictions like CPU , Clock, Cores and Memory and can not take advantage of computing from multiple compute nodes ?

Could you please advise as i had tried to create a VM as explained earlier of having 2GB ram of which both nodes have 6 and above GB ram memory and had tried using 2 or less cores but more GHz performance but still iam not able to utilize two nodes computing power, Could you please advise how to do this ?

Because if we can not create a big instance supporting good computing power how can we use CloudStack then for developing a private compute cloud ?

Thanks again.


Usman Malik MEMBERS
コメントアクション パーマリンク
Avatar
Pankaj Paliwal
Avatar

> I understand that CloudStack does not support CPU and Memory over-provisioning.
CPU over-provisioning is supported by CloudStack and all of the supported hypervisors. It allows you to run more VMs than the would otherwise be allowed and is controlled by cpu.overprovisioning.factor in Global Settings. This generally works fine unless most of the VMs are extremely busy, in which case there will be contention for CPU resources. CPU over-provisioning does not allow you to assign more CPU resources to a VM than is available on the host running the VM.

> This means that if it is showing "22 GHz" in infrastructure overview layout i can not create a big compute VM instance and would have to stick to the original host restrictions like CPU , Clock, Cores and Memory and can not take advantage of computing from multiple compute nodes ?
There is no way to assign the resources of multiple hosts to a single VM. This is not possible in any of the supported hypervisors. If you want to make use of the resources on multiple hosts for a specific application, create multiple VMs and run an application on those VMs that is capable of scaling across multiple servers. CloudStack includes load balancer functionality which can help with spreading network traffic across multiple VMs, or you can use your own load balancer to accomplish this.

Since your hosts have four cores, you won't be able to assign six or eight cores to a single VM since none of the hosts could run such a VM. Similarly, your hosts have 3.10 GHz CPUs so you won't be able to assign more than that to each core available to a VM.

Best regards,

{color:#555555}Kirk Kosinski{color} !http://www.linkedin.com/favicon.ico!
{color:#999999}MCITP: EA / VA / EDA7, VCP 4 / 5, CCA{color}


Kirk Kosinski CITRIX EMPLOYEES
コメントアクション パーマリンク

Top Contributors