Feature Server Cassandra Maintenance

The nodetool is a Cassandra utility for managing a Cassandra cluster. Use it for Feature Server Cassandra Maintenance.

Regular maintenance repairs inconsistencies across all data ranges, and ensures that replicated data is consistent across all nodes. You should perform maintenance on a node after a Feature Server upgrade and after new Feature Server nodes were added to the cluster. Apply additional maintenance steps after you remove a node from the cluster. See Performing maintenance after removing a node from the cluster for more details.

Performing regular maintenance

Run the nodetool command repair on all Feature Server hosts in a Cassandra cluster. See Running Nodetool on a Feature Server Host and Running nodetool repair on all Feature Server hosts in a Cassandra cluster for details about running nodetool and nodetool repair, respectively.

Important

All Feature Server hosts must be up and running to run repair. Running repair increases memory use and may not succeed if the Java heap space max limit is reached. For information on the Java heap size, see Hardware and software prerequisites.

As a best practice, schedule the regular maintenance weekly during the maintenance window (low usage hours), for example, by using Cron on Linux or as a Task Scheduler on Windows.

If you are planning for an ad hoc repair, then start the repair from the Feature Server instances that are configured as alternatevoicexml.

Running nodetool on a Feature Server host

The Nodetool utility has a command line interface. The steps below describe how to run nodetool commands on Cassandra when it is deployed on a host, as part of a Feature Server deployment.

1. Change the current directory.

In the Linux shell terminal:

cd <Feature Server Installation Directory>/work/jetty-<vms_host>-<http_port>-fs.war-_fs-any-/webapp/WEB-INF/lib

In the Windows command prompt window:

cd <Feature Server Installation Directory>\work\jetty-<vms_host>-<http_port>-fs.war-_fs-any-\webapp\WEB-INF\lib

where...

<Feature Server Installation Directory> is where Feature Server is installed in that host.

<vms_host> is the IP address of the vms host parameter that Feature Server started with, defined in launcher.xml or in the command line. The default is 0.0.0.0.

<http_port> is the port number that Feature Server started with, defined in launcher.xml. The default is 8080.

2. Run the Nodetool utility.

In Linux: (.jar is separated by a colon :)

java -cp libthrift-0.7.0.jar:cassandra-thrift-<Cassandra version>.jar:commons-cli-1.1.jar:cassandra-all-<Cassandra version>.jar org.apache.cassandra.tools.NodeCmd -h <Cassandra host> -p <jmx_port> <nodetool command>

In Windows: (.jar is separated by a semicolon ;)

java -cp libthrift-0.7.0.jar;cassandra-thrift-<Cassandra version>.jar;commons-cli-1.1.jar;cassandra-all-<Cassandra version>.jar org.apache.cassandra.tools.NodeCmd -h <Cassandra host> -p <jmx_port> <nodetool command>

where...

<Cassandra version> could be either 1.1.6 or 1.1.12 depending on Feature Server version installed. Please check the actual cassandra-all-*.jar file name in current directory.

<Cassandra host> is the hostname or IP address where Feature Server is running.

<jmx_port> is the JMX port number that Feature Server started with, defined in launcher.xml. Default=9192.

<nodetool command> can be ring [keyspace name] or removetoken <token> or repair or flush.

For example, the following command line runs nodetool and sends the command ring to the Feature Server running on the local host:

java -cp libthrift-0.7.0.jar:cassandra-thrift-1.1.12.jar:commons-cli-1.1.jar:cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 ring sipfs

Running nodetool repair on all Feature Server hosts in a Cassandra cluster

On Linux:

cd genesys/fs/work/jetty-0.0.0.0-8080-fs.war-_fs-any-/webapp/WEB-INF/lib

java -cp libthrift-0.7.0.jar:cassandra-thrift-1.1.12.jar:commons-cli-1.1.jar:cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 repair -pr

On Windows:

cd genesys\fs\work\jetty-0.0.0.0-8080-fs.war-_fs-any-\webapp\WEB-INF\lib

java -cp libthrift-0.7.0.jar;cassandra-thrift-1.1.12.jar;commons-cli-1.1.jar;cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 repair -pr

Performing maintenance after removing a node from the cluster

When you remove a node from the ring, you must also remove the corresponding tokens. Use these software procedures:

1. Run nodetool ring to obtain tokens. (See Running Nodetool on a Feature Server Host)
2. Run nodetool removetoken to remove nodes from the ring.
   Note: Only nodes that are down can be removed.
3. Run nodetool ring to validate removal.
4. Run nodetool repair on all Feature Server hosts in the cluster.
5. Run nodetool ring to validate repair.

For example, run nodetool ring to obtain tokens of nodes to be removed:

java -cp libthrift-0.7.0.jar:cassandra-thrift-1.1.12.jar:commons-cli-1.1.jar:cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 ring sipfs

…to see the following output returned tokens of nodes which are down:

<fs-host-IP1> DC1 RAC1 Up Normal 1.17 MB100.00% 167086018864645871692761019448293152722

<fs-host-IP2> DC1 RAC2 Up Normal 1.29 MB 100.00% 26003787676682001822918611294472056316

<fs-host-IP3> DC2 RAC1 Down Normal 1.15 MB 100.00% 41007983964572150951275225962045789866

<fs-host-IP4> DC2 RAC2 Down Normal 1.16 MB 100.00% 53685600614278234503162023330018045221

The following nodetool commands remove <fs-host-IP3> and <fs-host-IP4> nodes from the ring:

java -cp libthrift-0.7.0.jar:cassandra-thrift-1.1.12.jar:commons-cli-1.1.jar:cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 removetoken 41007983964572150951275225962045789866

java -cp libthrift-0.7.0.jar:cassandra-thrift-1.1.12.jar:commons-cli-1.1.jar:cassandra-all-1.1.12.jar org.apache.cassandra.tools.NodeCmd -h localhost -p 9192 removetoken 53685600614278234503162023330018045221

After token removal, running nodetool ring should provide the following output:

<fs-host-IP1> DC1 RAC1 Up Normal 1.17 MB 100.00% 167086018864645871692761019448293152722

<fs-host-IP2> DC1 RAC2 Up Normal 1.29 MB 100.00% 26003787676682001822918611294472056316

Scheduled Tasks

SIP Feature Server will now provide ability to add schedule tasks. These tasks should be programmed as Python script and executed in Feature Server environment by using Jython. All scheduled tasks should be configured in CME. All options intended for maintenance tasks are placed in section 'Scheduled Tasks'. Two predefined tasks can be scheduled to run periodically:

• Update-mailbox-counters
• Delete-expired-messages

Running scheduled tasks

Two ways to run Scheduled tasks: Run Schedule task manually by calling feature server resource: http://<fs-server-host:port>/fs/api/admin/tasks/start/<task-name> log in using administrative credentials to initiate scheduled task. OR Scheduled task started automatically by master feature server If option [Scheduled-task] <task-name>.active is set to true in master feature server application then scheduled task will run at time scheduled in option [Scheduled Tasks]<task-name>.schedule in master FS application

Cancelling scheduled tasks

Scheduled task can be cancelled by calling HTTP resource on master Feature Server:

http://<fs-server-host>:port/fs/api/admin/tasks/cancel/<task-name> 

Log in using your administrative credentials to cancel schedule tasks

Monitoring scheduled tasks

Feature Server has HTTP resource that will display the status of scheduled tasks Point your browser to this URL on the master Feature Server:

http://<fs-server-host>:<port>/fs/api/admin/tasks/state

Log in using your administrative credentials to display the current state of scheduled tasks, which can be one of these two:

• In progress
• Ready to start