Clusters and Nodes

Clusters and Nodes are the resources used to create and manage CipherTrust Manager clustering.

A cluster is a group of connected CipherTrust Manager appliances that share data.

The main purpose of clustering is to support High Availability. When clustered, all CipherTrust Manager appliances in the cluster (called cluster nodes or nodes) continuously synchronize their databases with each other - any appliance in the cluster may be contacted for any operation. A cluster supports any combination of appliance models; both virtual and physical appliances can join the same cluster.

If a member of the cluster gets temporarily disconnected, an alarm is set and syslog messages are sent. When the node comes back it synchronizes with other members of the cluster. If a member of the cluster gets permanently disconnected, other members should be notified by sending a ksctl cluster nodes delete command to one of them - the others will tell each other. This prevents the remaining nodes from continually storing catch-up changes for the missing node, which will eventually fill up their local volumes.

Nodes in a cluster communicate over port 5432. Members of the cluster must have bi-directional access to each other on port 5432.

Use the ksctl command line interface for managing clusters and nodes. The relevant commands are:

ksctl cluster csr
ksctl cluster delete
ksctl cluster fulljoin
ksctl cluster info
ksctl cluster join
ksctl cluster new
ksctl cluster nodes list
ksctl cluster nodes get
ksctl cluster nodes create
ksctl cluster nodes delete

What is not clustered?

Most things are clustered, however the following node specific items are not clustered, that is, are not replicated across the cluster.

• Backup files

• Backup keys

• Debug, KMIP Activity, and NAE Activity logs

• NTP configuration

• HSM configuration

• Instance name

• Virtual CipherTrust Manager license

• Interface certificates

• Proxy settings

Sharing HSM Root of Trust Configuration

CipherTrust Manager internally uses a flexible security architecture to allow clustering instances with both shared and distinct 'root of trust' configurations.

If the member node and the new node share the same HSM partition, it is recommended to pass the flag --shared-hsm-partition during nodes create or fulljoin, and to set each node to use the same Root of Trust (RoT) key using ksctl rot-keys rotate --id <root-of-trust-id>. This configuration further increases the security of cluster join procedure by ensuring that CipherTrust Manager secrets and master keys are protected by an HSM-secured RoT key during the join process, when these objects are transferred between nodes for the first time.

When the new node uses a distinct HSM partition or does not use any HSM, it is still possible to join it to an existing cluster by not specifying the --shared-hsm-partition flag.

Managing Clusters and Nodes

To check the status of a cluster

To check the status of a cluster enter the command:

$ ksctl cluster info

This returns the following response:

{
  "nodeID": "",
  "status": {
    "code": "none",
    "description": "not clustered"
  },
}

To create a new cluster

Make the following call on a running appliance (you will need to insert the hostname/IP of the appliance). When running cluster new, --host and --url are most likely going to be identical except that --host will not have the protocol. To create a new cluster enter the command:

$ ksctl cluster new --host=localHostName --url=urlOfCurrentNode --user=username --password=Password_1

This returns the following response:

{
  "nodeID": "ab40e178-5f1d-4f03-8b26-7ca378f74988",
  "status": {
    "code": "r",
    "description": "ready"
  },
  "nodeCount": 1
}

To show the status of all nodes in the cluster

Enter the following command on any node:

$ ksctl cluster nodes list

This returns the following response:

{
  "skip": 0,
  "limit": 256,
  "total": 1,
  "resources": [
    {
      "nodeID": "c05ead39-94ac-4459-a371-9915e3e16ebf",
      "status": {
        "code": "r",
        "description": "ready"
      },
      "host": "kylo_pg_1",
      "isThisNode": true
    }
  ]
}

The possible statuses are:

• nodes are in different states across databases

• not clustered

• ready

• joining: creating (1/5)

• joining: bootstrapping (2/5)

• joining: initial sync (3/5)

• joining: catching up (4/5)

• joining: completing (5/5)

• killed

• down

• unknown

To show the status of a single node in the cluster

To show the status of a single node in a cluster enter the command:

$ ksctl cluster nodes get --id=c05ead39-94ac-4459-a371-9915e3e16ebf

This returns the following response:

{
  "nodeID": "c05ead39-94ac-4459-a371-9915e3e16ebf",
  "status": {
    "code": "r",
    "description": "ready"
  },
  "host": "kylo_pg_1",
  "isThisNode": true
}

To join a node to a cluster

When joining a node to a cluster, three steps need to happen:

1. Get the csr from the joining node by entering the command ksctl cluster csr

2. Get the certificate and CA chain from a member node by entering the command ksctl cluster nodes create

3. Run the join command from the joining node by entering the command ksctl cluster join

   Caution

   The join operation might take some time to complete, depending on network speed and database size. Completion times of 30 minutes are not unusual.
   While a join is underway, do not restart the system, as this can cause the join to fail, and the node to disappear from the cluster. There is no need to manually restart the system after join is completed as well.
   You can always view the node status to check progress while a join is underway. During the join, the node displays a joining status and indicates progress through five stages: creating, bootstrapping, initial sync, catching up, and completing; the down status can also appear for a short time. If there is a failure during joining, the node status is killed or unknown, or a persistent down status.

4. Check that the join operation has completed and that the node status is ready before joining another node.

While all of these commands can be run individually, it is simpler to use the fulljoin command, which does all of these steps. For the cluster fulljoin command, you are required to input a member's IP or DNS, the joining node's IP or DNS, and either the joining node's configuration file, or its username, password, and URL. To complete the process using the fulljoin command enter the command:

$ ksctl cluster fulljoin --member=<member_IP_or_DNS> --newnodehost=<joining_IP_or_DNS> --newnodeconfig=<config_File_of_Joining_Node>.

As with the individual commands, you must wait until the join operation has completed and the node status is ready before joining another node.

An example of the cluster fulljoin command using a configuration file:

$ ksctl cluster fulljoin --member=memberIPOrDNS --newnodehost=joiningIPOrDNS --newnodeconfig=configFileOfJoiningNode

Response:

When you add a node to a cluster, the existing data of the node is deleted. Are you sure want to join? [y/N]

Attempting to get the CSR from the joining node...
Finished getting the CSR from the joining node...
Attempting to get the Certificate and CA Chain from the member node...
Finished getting the Certificate and CA Chain from the member node...
Attemping to join the new node to the member's cluster...

{
  "nodeID": "",
  "status": {
    "code": "creating",
    "description": "joining: creating (1/5)"
  }
}

When a node joins a cluster, the node adopts the credentials of the cluster. Would you like to write the new cluster credentials

to the provided configuration file? [y/N]

y

An example of the cluster fulljoin command without using a configuration file:

$ ksctl cluster fulljoin --member=memberIPOrDNS  --newnodehost=joiningIPOrDNS --newnodepass=joiningNodePass --newnodeuser=joiningNodeUser --newnodeurl=joiningNodeURL

Response:

When you add a node to a cluster, the existing data of the node is deleted. Are you sure want to join? [y/N]

y

Attempting to get the CSR from the joining node...
Finished getting the CSR from the joining node...
Attempting to get the Certificate and CA Chain from the member node...
Finished getting the Certificate and CA Chain from the member node...
Attemping to join the new node to the member's cluster...

{
  "nodeID": "",
  "status": {
    "code": "creating",
    "description": "joining: creating (1/5)"
  }
}

To remove a node from a cluster

To remove a node from a cluster, first remove that node from the cluster then delete the cluster configuration on the removed node. Deleting the cluster configuration allows a node to rejoin the original cluster, join another cluster, or create a new cluster.

1. Remove the node from the cluster. A node cannot remove itself, so you must call this on some other node in the cluster:

   $ ksctl cluster nodes delete --id=ebab0738-6e09-4b0d-8c99-850d7f24dfac
   

2. Delete the cluster configuration on the removed node:

   $ ksctl cluster delete
   

   Note

   If ksctl cluster delete doesn't work for any reason it is always possible to perform a full system reset to ensure any left over data is removed from the node. A node can be reset using the ksctl services reset command.

To rejoin a node to a cluster

Before rejoining it is recommended to reset the node.

To reset the node enter the command:

$ ksctl services reset

Join the node as described in to join a node to a cluster.

To create a new cluster from a removed node

Ensure the node have been successfully removed from the cluster and that ksctl cluster delete has been performed, as described in to remove a node from a cluster.

Create a new cluster as described in To create a new cluster.

Cluster Upgrade

A cluster can be upgraded in two ways:

• In-place Cluster Upgrade

• Using the Cluster Remove/Rebuild Method

In-place Cluster Upgrade

A cluster can be upgraded in-place since version 1.9.0. The upgrade is generally limited to one minor version at a time, for example, from 2.5.0 to 2.6.0; from 2.6.0 to 2.7.0; or from 2.7.0 to 2.8.0. Be aware of the following considerations when performing an in-place cluster upgrade.

• The node being upgraded will be inaccessible during the upgrade. This may be as long at 10 or more minutes. Clients must be able to handle this outage.

• There will be a brief period of time (under 30 seconds) where the database will be locked while upgrading the first node. This affects all nodes at the same time, and some nodes may give error responses during this time.

• All nodes in the cluster should be upgraded as soon as possible - nodes running different version of the firmware will behave differently, potentially causing problems with applications.

To perform an in-place cluster upgrade

1. Before doing any upgrade operation, ensure that you have a backup, and that you have downloaded the backup and associated backup key.

2. Ensure all nodes in the cluster are up and operating normally. Resolve any issues (like removing any obsolete nodes) before performing the upgrade.

3. Perform a system upgrade on each node, one at a time. Ensure the upgrade of each node is complete and that the node is operating normally, before proceeding to the next node.

   Note

   When updating the first node in a cluster, the cluster nodes may briefly experience slower than usual response times. This occurs because the shared database schema for the cluster is updated with the first node.

Upgrade using the cluster remove/rebuild method

1. On one of the cluster nodes, create and download a backup with corresponding backup key, in case there are any problems.

2. Remove all nodes from the cluster except one.

3. Delete the cluster configuration on the remaining node.

   $ ksctl cluster delete
   

4. Perform the upgrade on the remaining node.

   1. Ensure there is at least 12GB of space available (not including the upgrade file) before proceeding.

   2. scp the archive file to the CipherTrust Manager:

      $ scp -i <identity_file> <update file name> ksadmin@<ip>:.
      

   3. SSH into the CipherTrust Manager as ksadmin and run the following command:

      $ sudo /opt/keysecure/ks_upgrade.sh -f <~/filename>
      

   The signature of the archive file is verified and the upgrade is applied.

5. After the appliance reboots, re-build the cluster by creating a new cluster on this node.

6. Perform the upgrade on all other removed nodes.

   Note

   If a previously used node is to be re-used, the cluster must first be deleted from that system.

7. Join new instances to the cluster.