Rolling restart

This topic describes how to perform a rolling restart of the HDFS Transparency components in a Cloudera CDP managed environment.

Configuration updates to HDFS Transparency may be applied at runtime, without causing downtime to HDFS Transparency. This may be achieved by performing rolling restart of NameNodes and DataNodes as per the following logic:

Log in to an HDFS Transparency node as root and apply the configuration updates using mmhdfs config set command. However, if you are updating hadoop-env.sh, edit the /var/mmfs/hadoop/etc/hadoop/hadoop-env.sh file directly rather than using the mmhdfs config set command.
Upload the configuration to IBM Storage Scale CCR by running the mmhdfs config upload command.
Start with the rolling restarts of NameNodes. For rolling restart of NameNodes, follow the next procedure to make the NameNode state transitions quicker:
- Log in to one of the CES HDFS NameNodes as root.
- Find out which hosts are running the current active and standby NameNode, by running the following command:
```
/usr/lpp/mmfs/hadoop/bin/hdfs haadmin -getAllServiceState
```
- From Cloudera Manager > IBM Spectrum Scale service > instances, restart the currently Standby NameNode.
- Verify that the standby NameNode is running and is in good health. This should ensure there was no error with the applied configurations.
- On the active CES HDFS NameNode host as root, manually failover the active NameNode to the standby node (which was just restarted), by running following command:
```
/usr/lpp/mmfs/bin/mmces address move --ces-ip x.x.x.x --ces-node <standby_namenode_host>
```
- Verify and ensure that the active/standby nodes are now swapped, by running the following command as before:
```
/usr/lpp/mmfs/hadoop/bin/hdfs haadmin -getAllServiceState
```
- From Cloudera Manager > IBM Spectrum Scale service > instances, restart the other NameNode (which would now be in standby).
Continue with the rolling restart of DataNodes by following Rolling Restart in the Cloudera Manager documentation.