Adding compute profile types for auto scaling

Complete the following steps to add compute profile types when worker nodes are automatically added by the resource connector.

For users who only need to add multiple profiles, step 1 is sufficient and restart mbatchd is enough, The rest of steps are for submitting a job to a specific by tagging the vm with templateID as a string resource.

Add a new template section to the templates list. Review the values for all of the properties to ensure that they map to the correct VPC configuration. Update the file ibmcloudgen2_templates.json in $LSF_ENDIR/resource_connector/ibmcloudgen2/conf. A typical HPC cluster that is installed on IBM Cloud would have this file in /opt/ibm/lsf/conf/resource_connector/ibmcloudgen2/conf. See the following sample content for an example:

{
"templates": [
{
"templateId": "<TemplateID>",
"maxNumber": 200,
"attributes": {
"type": ["String", "X86_64"],
"ncores": ["Numeric", "1"],
"ncpus": ["Numeric", "1"],
"mem": ["Numeric", "1024"],
"icgen2host": ["Boolean", "1"],
"templateID": ["String", "<TemplateID>"]
},
"imageId": "<IMAGE_ID>",
"subnetId": "<SUBNET_ID>",
"vpcId": "<SUBNET_ID>",
"vmType": "cx2-2x4",
"securityGroupIds": ["<SECURITY_GROUP_ID>"],
"sshkey_id": "<SSH_KEY_ID>",
"region": "us-south",
"zone": "us-south-1",
"userData": "<TemplateID>"
}
]
}

Please use a valid value for Template ID. A '-' (hypen) cant be used in templateId when using it as a string resource. User may see this error: Error in select section: Operator "-" cannot be used with type , to get type . Job not submitted.

When the template is added, you can use the lsf.shared file to map the new template to a specific job. The lsf.shared file is located in $LSF_ENVDIR (same location as lsf.conf). In a typical installation, this would be found in the /opt/ibm/lsf/conf folder. Add templateId as a resource in the resource section in the lsf.shared file. For example, templateID String () () (template ID for the external hosts).

Add the following section to user_data.sh to let the virtual machines add templateId as LSF_LOCAL_RESOURCES.

if [ -n "$template_id" ]; then
sed -i "s/\(LSF_LOCAL_RESOURCES=.*\)\"/\1 [resourcemap $template_id*templateID]\"/" $LSF_CONF_FILE
echo "update LSF_LOCAL_RESOURCES in $LSF_CONF_FILE successfully, add [resourcemap ${template_id}*templateID]" >> $logfile
else
echo "templateID doesn't exist in environment variable" >> $logfile
fi

The user_data.sh is specified by the IBMCLOUDGEN2_PROVISION_FILE variable in the ibmcloudgen2_config.json file. ibmcloudgen2_config.json is located in /opt/ibm/lsf/conf/resource_connector/ibmcloudgen2/conf. See the following sample file content for an example:

"IBMCLOUDGEN2_KEY_FILE": "/opt/ibm/lsf/conf/resource_connector/ibmcloudgen2/credentials",
"IBMCLOUDGEN2_PROVISION_FILE": "/opt/ibm/lsf/conf/resource_connector/ibmcloudgen2/user_data.sh",
"IBMCLOUDGEN2_MACHINE_PREFIX": "icgen2host",
"LogLevel": "INFO"

Run the following commands to restart the processes and apply the changes:
```
$lsadmin reconfig
$badmin mbdrestart
```
You can submit the change by using the -R option. See the following example where "Template2" is an example value for Template ID:
```
bsub -R “templateId=Template2” sleep 1000
```