If you can't find the product or version you're looking for, visit support.ataccama.com/downloads

High Availability Component

The component acts as a Zookeeper Client that is part of each MDM node. It requires a Zookeeper Server to run. For an overview of Zookeeper, Zookeeper Overview.

The component actively communicates with Zookeeper to monitor the state of other MDM nodes and switches MDM between the RW and RO mode when necessary.

Required Components

HTTP Dispatcher
Server component (MDM)

Component Configuration

<component class="com.ataccama.nme.ha.HighAvailabilityComponent">
    <serverName>mdc1</serverName>
    <serverHostname>localhost</serverHostname>
    <zookeeperPath>/ataccama/MDM_ha_default</zookeeperPath>
    <zookeeperHostname>localhost:2181</zookeeperHostname>
    <databaseLogger>db_logger</databaseLogger>
    <consoleListenerName>ha-console</consoleListenerName>
    <revokeLeadershipOnDisconnectTimeout>15000</revokeLeadershipOnDisconnectTimeout>
    <maxActiveToPassiveSwitchingTime>15000</maxActiveToPassiveSwitchingTime>
 </component>

Parameter Name Required Description

serverName

Yes

Name of this MDM shown in the Zookeeper cluster status.

Each MDM must have a unique name, for example, MDM1, MDM2 and so on.

serverHostname

Yes

Hostname by which other MDMs can reach this instance.

zookeeperPath

Path on the zookeeper file system. Default is /ataccama/mdc_ha_default.

zookeeperHostname

Yes

Zookeeper cluster hostname. Multiple hostnames can be separated by a comma: localhost:2181,localhost:2182,localhost:2183.

databaseLogger

name of the database logger (configured in Logging Configuration) to show logs from an MDM node.

consoleListenerName

Name of the HTTP Dispatcher listener on which the HA Admin Console will be available. You need to configure a dedicated Jetty listener first.

revokeLeadershipOnDisconnectTimeout

Yes

When a connection failure occurs, this specifies how long Zookeeper waits to revoke leadership from the disconnected client. If the disconnected node reconnects during the timeout, it maintains its leadership. If the disconnected node fails to reconnect, Zookeeper selects a new leader and the disconnected node starts switching to the RO mode. Basically, this is session timeout and connection timeout in terms of Zookeeper documentation. Default: 15 000 ms.

This timeout must be the same across all MDM instances.

maxActiveToPassiveSwitchingTime

Yes

The maximum amount of time it might take for an RW node to switch to the RO mode (for whatever reason). When an RW node cannot switch to the RO mode fast enough, it is killed. A new leader (and soon-to-become RW node) waits this time + 3 s until it starts switching to the RW mode. This is a counter-measure against having two or more RW nodes in the cluster at the same time. Default: 10 000 ms.

This timeout must be the same across all MDM instances.

Was this page useful?