ClustrixDB requires a number of ports to allow internode communication. Depending on your specific deployment platform, you can:
There are 9 ports that are used for communication between ClustrixDB nodes. These ports are:
Additional ports are required when the multiport feature is enabled (see below)
The multiport feature was introduced in ClustrixDB Release 7.0 and allows internode communication from any core in the cluster. This feature can greatly increase performance under heavy load, but also requires more open ports. Without multiport enabled, all messaging is handled through core 0 of each node.
For example, a 3 node cluster with 16 cores/hyperthreads per node you would have a total of 48 core (3 nodes times 16 core), so you should open ports 24379 - 24427.
As you add more nodes, this port range also needs to be be expanded.
Alternatively, you can disable Multiport and only open the 9 ports listed above. (not recommended)
If you are using a firewall (like iptables) between the nodes, please open the specific ports.
Multiport is enabled by default in v7.0 and later releases. Please see Modifying Startup Configuration Options for instructions on how to disable multiport.
ClustrixDB does not support mixed-mode usage of multiport; all nodes must uniformly have multiport enabled (default) or disabled, or the nodes will be unable to form a cluster.
This section describes how to enable the nodes to connect to each other without a password. This is required for:
The installer will set up host based authentication (ON by default) and you can easily setup key pair authentication if that is more desired.
During the installer, choose option 11 and toggle to Yes
This will enable Host Based Authentication between the nodes allowing them to communicate with each other via ssh. The default for this is 'Yes'.
First generate the key pair. You'll want to keep the file names default as this is what clx expects. If you have an existing key pair you can either skip this first step, or rename it to something other than id_rsa.
You will see output like:
Now you need to create/append to authorized keys and then copy the files to the other nodes in the cluster. The below example assumes there is a root password, foo specified with -R flag. If there is no root password on the nodes at the time these steps are run, the -R flag does not need to be used.
Run the following commands:
That's it! You can test this with:
If dates are correctly returned, everything is working as expected.
As long as all the nodes have the same password, you can run clx with the -R command. Once clx has been run with the -R command, this will be cached and upgrades and other clx operations will function as expected.