How to Start a SnappyData Cluster
Starting SnappyData Cluster on a Single Machine
If you have downloaded and extracted the SnappyData product distribution, navigate to the SnappyData product root directory.
Start the Cluster: Run the
./sbin/snappy-start-all.sh script to start the SnappyData cluster on your single machine using default settings. This starts a lead node, a locator, and a data server. The Hive Thrift server also starts by default.
It may take 30 seconds or more to bootstrap the entire cluster on your local machine. An additional 10 seconds is required to start the Hive Thrift server. To avoid this additional 10 seconds, you can set the
snappydata.hiveServer.enabled to false.
Sample Output: The sample output for
snappy-start-all.sh is displayed as:
Logs generated in /home/cbhatt/snappydata-1.1.1-bin/work/localhost-locator-1/snappylocator.log SnappyData Locator pid: 10813 status: running Distributed system now has 1 members. Started Thrift locator (Compact Protocol) on: localhost/127.0.0.1 Logs generated in /home/cbhatt/snappydata-1.1.1-bin/work/localhost-server-1/snappyserver.log SnappyData Server pid: 11018 status: running Distributed system now has 2 members. Started Thrift server (Compact Protocol) on: localhost/127.0.0.1 Logs generated in /home/cbhatt/snappydata-1.1.1-bin/work/localhost-lead-1/snappyleader.log SnappyData Leader pid: 11213 status: running Distributed system now has 3 members. Starting hive thrift server (session=snappy) Starting job server on: 0.0.0.0
Starting the SnappyData Cluster on Multiple Hosts
To start the cluster on multiple hosts:
The easiest way to run SnappyData on multiple nodes is to use a shared file system such as NFS on all the nodes. You can also extract the product distribution on each node of the cluster. If all nodes have NFS access, install SnappyData on any one of the nodes.
Create the configuration files using the templates provided in the conf folder. Copy the existing template files (servers.template, locators.template and leads.template) and rename them to servers, locators, leads. Edit the files to include the hostnames on which to start the server, locator, and lead. Refer to the configuration section for more information on properties.
Start the cluster using
./sbin/snappy-start-all.sh. SnappyData starts the cluster using SSH.
Starting Individual Components
Instead of starting SnappyData cluster using the
snappy-start-all.sh script, individual components can be started on a system locally using the following commands:
All configuration parameters are provided as command line arguments rather than reading from a configuration file.
For example, you can run any of the following commands depending on the individual component that you want to start.
$ ./bin/snappy locator start -dir=/node-a/locator1 $ ./bin/snappy server start -dir=/node-b/server1 -locators=localhost -heap-size=16g $ ./bin/snappy leader start -dir=/node-c/lead1 -locators=localhost -spark.executor.cores=32
The path mentioned for
-dir should exist. Otherwise, the command will fail with FileNotFoundException.
Executing Commands on Selected Cluster Component
cluster-util.sh (--on-locators|--on-servers|--on-leads|--on-all) [-y] (--copy-conf | --run "<cmd-to-run-on-selected-nodes>")
You can use the cluster-util.sh utility to execute a given command on selected members of the cluster. The script relies on the entries you specify in locators, servers, and leads files in the conf directory to identify the members of the cluster.
-on-locatorsIf specified, the given command is executed on locators.
--on-serversIf specified, the given command is executed on servers.
--on-leadsIf specified, the given command is executed on leads.
--on-allIf specified, the given command is executed on all the member in the cluster.
-yIf specified, the script does not prompt for the confirmation to execute the command on each member node.
--copy-confThis is a shortcut command. When you specify this comand, the log4j.properties, snappy-env.sh and spark-env.sh configuration files are copied from the local machine to all the members. These files are copied only in the following conditions:
- If these are absent in the destination member
- If their content is different. In latter case, a backup of the file is taken in conf/backup directory, on the destination member, before copy.
--run <cmd-to-run-on-selected-nodes>If specified, the given command(s) is executed on specified members. Command to be executed specified after --run`` must be in double-quotes.
1.To copy configuration files on all servers “./sbin/cluster-util.sh --on-servers --copy-conf 2.To run “ls” command on all servers with -y option “./sbin/cluster-util.sh --on-servers -y --run “ls”