Originally posted by: Jenna Lau-Caruso

Starting with IBM Spectrum Conductor 2.4.0, you can configure a cluster installed on a local file system, to deploy application instances and instance groups to a shared file system.
 

This configuration allows your to install your cluster's management services to local disk, increasing the performance of these critical services, while still taking advantage of time-savings and space-efficiencies of deploying application instances, Anaconda distributions and instance groups to a shared file system.
 

To set up shared deployments on a cluster which has been installed to a local file system, perform these simple configuration steps:

1. Configure the ELASTIC_HARVEST_LOCATION environment variable
   • To configure at installation time, specify the ELASTIC_HARVEST_LOCATION environment variable to a location on a shared file system that is accessible by all hosts in the cluster.
   • To configure if you have already installed IBM Spectrum Conductor, modify the ELASTIC_HARVEST_LOCATION environment variable in the elk-shipper and SparkCleanup services:
     <ego:EnvironmentVariable name="ELASTIC_HARVEST_LOCATION">

/path/on/shared/filesystem

</ego:EnvironmentVariable>

Restart the services for the changes to take effect.

2. Configure the ascd service for shared deployments
   1. Edit the $EGO_CONFDIR/../../ascd/conf/ascd.conf file to set ASCD_SHARED_FS_DEPLOY=ON.
   2. Restart the ascd service for the change to take effect.
       
3. Optionally configure the SparkCleanup service to run on a single host
   1. Make the following changes to the SparkCleanup service to configure the service to run on a single host:
      <sc:MaxInstances>1</sc:MaxInstances>

<ego:ConsumerID>/ManagementServices/EGOManagementServices</ego:ConsumerID>

<ego:ResourceGroup>ManagementHosts</ego:ResourceGroup>

2. Restart the service for the changes to take effect.

 Note: Skip this step if you plan to enable the shuffle service on your instance groups, specifying a local file system location for the shuffle data.

Once configured, ensure that you specify a location on a shared file system for all application instances, Anaconda distributions, and instance groups. A single host will automatically be selected for each package deployment.

Read more about it!

For more information about installing IBM Spectrum Conductor on a local file system with shared file system deployments, check out the detailed steps in IBM Knowledge Center. 

Let us know what you think!

Give it a try and share your feedback on our Slack channel.