Removing a Node from an On-premise Fabric
Describes how to remove one or more nodes from an on-premise fabric deployment by using the Data Fabric UI.
Removing nodes from an installed fabric is necessary in many situations such as:
- To remove the nodes that have reached end of life.
- To perform maintenance activity such as applying security patches or upgrading the operating system.
- To perform some activities mandated by organization that need node removal.
- To scale down a fabric that was scaled up to meet a temporary requirement of increased storage and compute.
Prerequisites
IMPORTANT
To avoid loss of data, remember to back up the data from the node,
before you remove the node.For on-prem fabrics, you can choose a node to remove. Ensure that the node satisfies the criteria mentioned in this section.
Following are the prerequisites for node removal.
- You cannot use the Data Fabric UI to remove nodes from a cluster that was installed using a method other than the seed node installation method.
- You must have fabric manager credentials to remove a node.
- A fabric requires a minimum of two CLDB and Zookeeper hosts at all times. A node
cannot be removed if this condition is breached due to the node removal.NOTEIf you remove a control node (that have CLDB and ZooKeeper running), cluster will be reconfigured with the remaining CLDBs, and Warden will be restarted on each node of the fabric, after removing the control node. Before proceeding to remove a CLDB master node, perform failover of CLDB master first, and then proceed with node removal.
- If the fabric running the node is the primary fabric in the global namespace, you cannot remove the node that runs the SSO Keycloak service.
- You cannot remove the node that has the installer host that has been used to configure the fabric or the current node that is being used for node removal.
- You can remove a node only if the fabric has at least three nodes, excluding the node or nodes to be removed.
- You can remove a node only if the cluster storage consumption is less than 80 percent.
- On any given fabric, only one Remove node operation can be done at a time. If a remove node operation is already in progress, you cannot initiate another add node or remove node operation on the same fabric. However, multiple remove node operations can be performed at the same time on different fabrics belonging to the same global namespace.
- Remove node operation is not possible when an upgrade fabric, remove fabric, or add node operation is already in progress.
Steps for Removing Nodes from a Fabric
Use the following steps to remove nodes:
- Log on to the Data Fabric UI.
- Select the Fabric manager option.
- On the Global namespace card, check the table view.
- Under the Resource Name column, click the name of the fabric for which you want to remove nodes. The Fabric details are displayed.
- Click the Nodes tab. For more information, see Viewing Node Information.
- Click the ellipsis (
) in the Action column, and select Delete . The Delete node side drawer appears.
- Fill in the following node parameters:
For this parameter . . . Do this SSH credentials* Enter your Username for SSH access to the node. Enter your Password.
- Optionally, turn on the Skip host drain check toggle, to skip checking if the data on the node has been drained before removing it, after three failed attempts to remove the node. The node drain check is skipped the fourth time, if there are three failed node removal attempts, and the node is removed without checking if the data from the node to be removed has been drained (that is, moved to another node on the same cluster).
- Click Delete.
Results:
The Data Fabric UI displays Deleting in Status column for the node that is being deleted.If the node removal/delete operation is successful, the Data Fabric UI displays a success message, and the table of active nodes is updated.