Hi All,

I am planning to install 2 MAS instances on same openshift cluster with 3 worker nodes. Is their a way to taint or label the nodes so that one of my MAS instances and it related workloads are deployed on that particular nodes and not the others?

I am aware this is something which can be done for DB2U namespace. I want to ensure that all my MAS instance related workloads reside on a particular node? Is their a way to achieve this, is their a config option for this during mas-install?

Thanks

Abhinav

Hi Abhinav,

there are two ways You can do that. First one is to use MAS podTemplates functionality, but You would need to put in all the pods in their respective CRs, which manage them. By assigning affinity You can specify which pod should run on which node. More: https://www.ibm.com/docs/en/masv-and-l/cd?topic=workloads-customizing-workload-affinity

Second option is OpenShift native and that is the one I would use. Put a label on the target node (for example nodeUse=masdev ) and in the MAS namespace put in this annotation

openshift.io/node-selector: nodeUse=masdev

This will cause the OpenShift scheduler to place any new pod on the node matching the selector. It only applies to the newly created pods, so if You have those pods already running somewhere You would need to restart them.

Hi All,

I am planning to install 2 MAS instances on same openshift cluster with 3 worker nodes. Is their a way to taint or label the nodes so that one of my MAS instances and it related workloads are deployed on that particular nodes and not the others?

I am aware this is something which can be done for DB2U namespace. I want to ensure that all my MAS instance related workloads reside on a particular node? Is their a way to achieve this, is their a config option for this during mas-install?

Thanks

Abhinav

Hi Witold,

Thanks for the reply. I will try the 2nd approach which you mentioned, just a quick question since MAS install creates multiple namespaces for example mongoce, db2u, grafana5, mas-{mas-instance-id}-core,mas-{mas-instance-id}-manage, mas-{mas-instance-id}-pipelines, mas-{mas-instance-id}-sls. So all these should be labelled on the node correct?

And what about the namespaces mongoce, db2u, grafana5 since these do not get created with a instance identifier, any way to distinguish them across multiple mas instance deployments and apply node labelling to them?

Abhinav

Hi Abhinav,

there are two ways You can do that. First one is to use MAS podTemplates functionality, but You would need to put in all the pods in their respective CRs, which manage them. By assigning affinity You can specify which pod should run on which node. More: https://www.ibm.com/docs/en/masv-and-l/cd?topic=workloads-customizing-workload-affinity

Second option is OpenShift native and that is the one I would use. Put a label on the target node (for example nodeUse=masdev ) and in the MAS namespace put in this annotation

openshift.io/node-selector: nodeUse=masdev

This will cause the OpenShift scheduler to place any new pod on the node matching the selector. It only applies to the newly created pods, so if You have those pods already running somewhere You would need to restart them.

Hi All,

I am planning to install 2 MAS instances on same openshift cluster with 3 worker nodes. Is their a way to taint or label the nodes so that one of my MAS instances and it related workloads are deployed on that particular nodes and not the others?

I am aware this is something which can be done for DB2U namespace. I want to ensure that all my MAS instance related workloads reside on a particular node? Is their a way to achieve this, is their a config option for this during mas-install?

Thanks

Abhinav