Snitch determines which datacenters and racks nodes belong to, and Snitches inform Scylla about the network topology so that requests are routed efficiently and allows Scylla to distribute replicas by grouping machines into data centers and racks.
Note that if you do not choose a Snitch when creating a Scylla cluster, the SimpleSnitch is used by default.
Scylla supports the following snitches:
GoogleCloudSnitch (New in Scylla 2.2)
For production clusters, it is strongly recommended to use
Other snitches are limited and will make it harder for you to add a Data Center (DC) later.
When working with single cluster deployments and all the nodes are under the same datacenter, use the SimpleSnitch, it binds all the nodes to the same Rack and datacenter. This Snitch is used in a single datacenter deployments only.
Bind Nodes to DCs and Racks according to their Broadcast IPs:
Node has a Broadcast IP 192.168.100.200; than it would belong to a DC ‘168’ and Rack ‘100’
When working with multi-cluster deployments and the nodes are in various datacenters, use the GossipingPropertyFileSnitch.
It is recommended to be used in production installations. It allows us to explicitly define which DC and Rack a specific Node belongs to.
This Snitch reads its configuration from a
cassandra-rackdc.properties file that located under
prefer_local=true dc=my_data_center rack=my_rack
prefer_local to true will tell Scylla to use a local IP address. The local IP address is the listen_address parameter from the scylla.yaml file which is located under
Use this Snitch when working on EC2 with single cluster deployments where all nodes are under the same region. This basic Snitch reads its configuration from Amazon’s EC2 registry services. In EC2, the region name is treated as the datacenter name, and availability zones are treated as racks within a datacenter. If the setup includes a single datacenter, there is no need to specify any parameters. Because private IPs are used, this Snitch does not work across multiple regions. It should also be noted that according to this Snitch, a DC is a region. If a region is down, the entire cluster will be down.
If you are working with multiple datacenters, specify the DC and set the parameter
dc_suffix=<DCNAME> in the
cassandra-rackdc.properties file under
For Example, suppose you had created a 5 node cluster and added the following to each node’s /etc/scylla/cassandra-rackdc.properties file as shown:
Parameter to add to the specific node’s /etc/scylla/cassandra-rackdc.properties
This action adds a suffix to the name of each of the datacenters for the region. Running nodetool status command shows all three datacenters:
Datacenter: us-east_dc1-europe ============================== Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN 172.20.0.4 111.23 KB 256 ? eaabc5db-61ff-419b-b1a7-f70af23edb1b Rack1 UN 172.20.0.5 127.09 KB 256 ? bace1b4e-67c6-4bdb-8eba-398162b7b56e Rack1 Datacenter: us-east_dc2-asia ============================ Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN 172.20.0.6 110.59 KB 256 ? bda5fb11-9369-48fb-91be-82c8d821f758 Rack1 UN 172.20.0.3 111.26 KB 256 ? b9ea3516-5e1e-4ffb-abff-c6a6701cb41b Rack1 Datacenter: us-east_dc3-australia ================================= Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN 172.20.0.7 111.23 KB 256 ? eaabc5db-61ff-419b-b1a7-f70af23edb1b Rack1
The datacenter naming convention in this example is based on location. You can use other conventions, such as DC1, DC2 or 100, 200, or analytics, search, Scylla, and more. Providing a separator such as a dash keeps the name of the DC readable as the
dc_suffix property adds the suffix to the DC name.
Ec2Snitch and Ec2MultiRegionSnitch will define DC/RACK differently for AWS Availability Zones (AZs) that end with
1x compared to other AZs:
For the former class of AZs, e.g.
us-east-1d, the Snitch will set
For the latter class of AZs, e.g.
us-east-4c, the Snitch will set
When working on EC2 using multi-cluster deployments and the nodes are in various regions, use the Ec2MultiRegionSnitch.
Works like the Ec2Snitch, in addition, it set the
broadcast_rpc_address of a node to the node’s public IP.
This would allow nodes from other zones to communicate with this node regardless of what is configured in the node’s scylla.yaml configuration file for
Ec2MultiRegionSnitch also unconditionally imposes the “prefer local” policy on a node (similar to GossipingPropertyFileSnitch when prefer_local is set to true).
In EC2, the region name is treated as the datacenter name and availability zones are treated as racks within a datacenter.
To change the DC and Rack names, do the following:
cassandra-rackdc.properties file with the preferred datacenter name. The file can be found under
dc_suffix defines a suffix added to the datacenter name as described below.
Node - region
DC='us-west' and Rack
Rack='1' will be
Node - region
DC='us-east' and Rack
Rack='2' will be
Use the GoogleCloudSnitch (available in Scylla version 2.2 and later) for deploying Scylla on the Google Cloud Engine (GCE) platform across one or more regions. The region is treated as a datacenter, and the availability zones are treated as racks within the datacenter. All communication occurs over private IP addresses within the same logical network.
To use the GoogleCloudSnitch, add the Snitch to the scylla.yaml file, which is located under
/etc/scylla/ for all nodes in the cluster.
You can add a suffix to the data center name as an additional identifier. This suffix is appended to the Zone name without adding any spaces. To add this suffix edit the
cassandra-rackdc.properties file, which can be found under
/etc/scylla/ and set the
dc_suffix with an appropriate text string. It may help to add an underscore or dash in front. Keep in mind that this property file is used for all Scylla snitches. When using GoogleCloudSnitch, all other properties are ignored.
You have two datacenters running on GCE. One is located in region us-east1, and zone us-east-1-b and is allocated for your office in Miami. The other is located in us-west1 and zone us-west-1-b and is allocated for your office in Portland. In this case, the data center one’s name is us-east1 with rack name b, and the second DC is us-west1 with rack b. (Racks are important for distributing replicas, but not for datacenter naming.) This Snitch can work across multiple regions without additional configuration.
After creating the instances on GCE, edit the scylla.yaml file to indicate the GoogleCloudSnitch. While you have the file open, check that you have at least two seed nodes, one in each datacenter.
As you want to set the data center suffix for the nodes in each datacenter, you open each node’s properties file in the
cassandra-rackdc.properties The file can be found under
/etc/scylla/. You set the following parameters for Miami:
# node 1 - 192.0.2.2 (you use the same properties for node #2 (192.0.2.3) and #3 (192.0.2.4)) dc_suffix=_scylla_node_Miami
and for Portland:
# node 4 192.0.2.5 dc_suffix=_scylla_node_Portland
You start the cluster, one node at a time, starting with the seed nodes, and then run
nodetool status to check connectivity.
nodetool status Datacenter: us-east1_scylla_node_Miami ====================================== Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN 192.0.2.2 1.27 MB 256 ? 5b1d864f-a026-4076-bb19-3e7dd693abf1 b UN 192.0.2.3 954.89 KB 256 ? 783a815e-6e9d-4ab5-a092-bbf15fd76a9f b UN 192.0.2.4 1.02 MB 256 ? 1edf5b52-6ae3-41c1-9ec1-c431d34a1aa1 b Datacenter: us-west1_scylla_node_Portland ====================================== Status=Up/Down |/ State=Normal/Leaving/Joining/Moving -- Address Load Tokens Owns Host ID Rack UN 192.0.2.5 670.16 KB 256 ? f0a44a49-0035-4146-8fdc-30e66c037f95 b