Skip to content

img0

Uninstall Node

Uninstalling a Node

Warning

When uninstalling a node, the number of remaining nodes must be sufficient to meet the needs of your existing Storage Policies.

Instructions

  1. Login as Admin user
  2. Click Cluster Tab
  3. Click Storage Policies Sub-Tab
  4. Ensure you delete any policies you have created that require more than nodes than the number of nodes that will be left (e.g. EC 3 + 2)
  5. You may need to remove Objects and Buckets beforehand if you have created anything using these Policies.

Information

Prior to removing the node it is best practice to put the node into maintenance mode. This ensures that:
The node will not generate alerts
If the node is hosting a Credentials DB master or QoS DB master, the master role will be temporarily shifted to a different node
The S3 Service on other nodes will not direct requests to the HyperStore Service or the Metadata DB on the node that is maintenance mode

Instructions

  1. Select Cluster Tab
  2. Select Nodes Sub-Tab
  3. Select Advanced
  4. Select Start Maintenance Mode
  5. Select Target Node
  6. Click Execute

Instructions

  1. From the main Cluster tab, confirm the node is demonstrated as in Maintenance Mode with a blue icon replacing the green icon.

Information

When the node is in maintenance mode, we should also stop all requests from being sent to it. The easiest way to do this is by draining the node in HyperBalance.

Instructions

  1. Login to your Primary HyperBalance using User: loadbalancer and the password provided by your trainer
  2. Expand the view of one Virtual Service in the System Overview page
  3. Left click on the word Drain on the row with your node 5. a popup box appears and asks for confirmation. Click OK to continue
  4. A second popup box appears, asking you to confirm drain of all services registered in the HyperBalance for node 5. Click OK to confirm

Information

When the node has been drained completely node 5 will show as Online (Drain) with a status of None

Info

If you are removing a node after having added a new node to your cluster, you must complete the rebalance operation for the new node before removing a node.
If the node you removed was "dead", meaning the Cassandra Service on the node is down or unreachable when you removed the node you MUST repair the remaining nodes in your cluster (this is not necessary if the node you removed was live).

Info

You must ensure you have sufficient available storage capacity on the other nodes in your cluster. If any nodes (or disks within nodes) are in a “stop-write” condition, the decommissioning process overrides this restriction and the disk may become full, resulting in stream job failures once the disks can accept no more data.

Warning

Removing a node should be something that you do only if absolutely necessary. When you remove a live node from your cluster, during the decommissioning process the data replicas and erasure coded fragments on that node are unavailable to help meet your configured read consistency requirements, when servicing read requests from client applications.

Once data has been streamed from the decommissioning node to the other nodes that data is again available to contribute to meeting read consistency requirements.

Depending on the data volume and network bandwidth, it may take several days or more until the decommissioning process has completed for all of the node's data.

Instructions

  1. Select Cluster Tab
  2. Select Nodes Sub-Tab
  3. Select Advanced
  4. Select Uninstall Node
  5. Select Target Node (node5)
  6. Click Execute
  7. Click OK in the popup box to confirm that you want to run the Uninstall Node command

Info

Consider if you wish to delete data and logs at the same time by selecting the check box

Uninstall a Node

Info

Once Uninstall operation completes, check Cluster Tab – Data Centres and ensure you now only have 4 Nodes configured.