Setting the amount of memory per partition for Db2 Warehouse MPP

You can specify the amount of memory to allocate per partition, or multiple logical node (MLN), in a Db2® Warehouse massively parallel processing (MPP) configuration by editing the JSON file that is used by the service.

About this task

By default, a Db2 Warehouse service on Cloud Pak for Data:

Uses 24 GiB of memory per worker node, considered a minimal allocation for testing. You can change this allocation in the web console when you deploy a Db2 Warehouse service. Note that the web console unit for memory is GiB. To convert GiB to GB, multiply the GiB value by 1.048.
Uses 24 GiB of memory per partition (MLN). You cannot change this allocation in the web console, but you can change it by following this procedure.

When you specify a memory allocation per MLN, the database deployment process looks at the total available memory across all worker nodes and creates a number of MLNs by dividing total memory by memory-per-MLN. For example, if your configuration uses three worker nodes with a memory allocation of 64 GB per worker node, a total of 192 GB of memory is available for the configuration. If you want four MLNs for the configuration, you would edit the JSON file to specify memory-per-MLN of 48 GB.

For help with determining your memory allocation and the optimal number of partitions to create, you can use the Db2 Warehouse online configuration tool.

Note: You must make any changes to the JSON file before you deploy the database.

Procedure

Locate the zen-database-core pod:
```
zdc=$(oc get po | grep zen-database-core | awk {'print $1'}) 
```
Where zdc is the name that will be returned for the zen-database-core pod. You will use this value in Step 2.
Run the exec command to access an interactive shell inside of the container:
```
oc exec -it ${zdc} bash 
```
Change to the /user-home/_global_/databases directory.

Run the following statements in the bash shell:

jf=$(ls -1 /user-home/_global_/databases/ibm-db2wh-*-x86_64.json)
python <<EOF
import json
jf = '$jf'

with open(jf, 'r') as fd:
    parsed = json.load(fd)

parsed["create"]["mem-per-mln"]=memory_per_mln

with open(jf, 'w') as fd:
    json.dump(parsed, fd, indent=2)
EOF

Where memory_per_mln is the amount of memory that you want to allocate to each MLN.

The statements will update the ibm-db2wh-version-x86_64.json file to set the default memory allocation per node.