Resource planing

Nov 3, 2010 at 10:08 PM

Hello,

I have two questions on resource planing in Trident:

1. In the run/schedule workflow dialog of Trident Management Studio it is possible to select the node to run the workflow. What do you have to select in case of HPC Cluster the cluster node or the cluster head to run the workflow?

2. In case there is HPC Cluster with two cluster heads and one of them down, will Trident scheduler schedule workflows over the cluster head which is alive, or is the IP of the cluster head "hard configured"?

 

Thanks,

Alex

 

 

Nov 4, 2010 at 3:03 PM

Hi Alex,
 
We are looking into your question. We will keep you posted on the updates.
 
Please let us know if you have any further questions.
Regards,
Trident Support Team

Nov 10, 2010 at 12:59 PM

Hi Alex,

Thanks for your patience. We are working on your question with the highest priority.

We will keep you posted on the updates.

Please let us know if you have any further questions.
Regards,
Trident Support Team

Nov 24, 2010 at 2:27 PM

Hi,

We are currently testing these scenarios on the HPC Clusters. We will post the response to your questions at the earliest. 

Regards,
Trident Support Team   

Dec 1, 2010 at 12:18 PM

Hi,
 
Thanks for your patience. Please find the response to your questions below:
 
<<What do you have to select in case of HPC Cluster the cluster node or the cluster head to run the workflow>>
 Trident Workflows are executed in the HPC Environment using the Trident LightWeightExecutor. Trident LightWeightExecutor is a console application, which accepts the Id of the workflow to be executed and an optional input of path to the file where the workflow inputs are specified
 
e.g. LightWeightExecutor.exe –wf <<WorkflowID>> <<OptionalInputPath>>
 
Trident uses HPC Scheduler to run the jobs. We schedule the jobs on the cluster’s head node. The actual workflow will run on the cluster’s compute nodes.
 
<< In case there is HPC Cluster with two cluster heads and one of them down, will Trident scheduler schedule workflows over the cluster head which is alive, or is the IP of the cluster head "hard configured">>
 We first install Trident on the cluster’s head node. The head node information is stored in the Trident registry.
 
Important:
1.       Make sure the compute nodes are able to access Head node SQL server with the authentication method you will be using with installer.
2.       Trident should be configured on Head node before installing compute node.
3.       User installing Trident on compute node should be a valid user on Trident registry
 
After installing Trident on the cluster’s head node, we install it on the cluster’s compute node. The compute node points to the head node information stored in the Trident registry. Trident’s runtime will use the head node information in the registry to submit jobs to HPC scheduler.  If the head node is down, then there will be an error.
 
Please let us know if this helps in answering your questions.
Regards,
Trident Support Team

Dec 2, 2010 at 9:57 PM

Thanks a lot!

It answered my questions.