Configuring HA for TimelineServer

Scenario

As a role of the Yarn service, TimelineServer supports the HA mode since the current version. To prevent a single point of failure of TimelineServer, you can enable TimelineServer HA to ensure high availability of the TimelineServer role.

The TimelineServer saves data to the in-memory database LevelDB, which is memory-intensive. It is imperative to allocate a minimum of 30 GB of memory to the node hosting the TimelineServer.

This function applies to MRS 3.2.0-LTS.1 or later.

Impact on the System

• Before the conversion, change the value of TLS_FLOAT_IP of TimelineServer to an available floating IP address. (In the case of a single instance, the service IP address of the node is used by default.)
• During the conversion, the configuration of the TimelineServer role will expire. In this case, you need to restart the instance whose configuration expires.

Procedure

1. Log in to FusionInsight Manager and choose Cluster > Services > Yarn. Click Configurations.
2. Change the value of TLS_FLOAT_IP to an available floating IP address (the floating IP address and the service IP addresses of the two TimelineServer instances must be in the same network segment) and click Save then OK.
3. Click the Instance tab. Click Add Instance, select a node to add a TimelineServer instance, and choose Next > Next > Submit. The instance is added.
4. On FusionInsight Manager, click next to the cluster name, select Restart Configuration-Expired Instances, and wait until the instance is restarted.
5. Check the status of each instance after the restart. For example, the active/standby status and running status of the TimelineServer instances are normal.