Products | Versions |
---|---|
Spotfire Data Science | 6.x |
Connecting Alpine to a cluster with Resource Manager High Availability enabled
Connecting Alpine to a cluster with Resource Manager High Availability enabled
If Resource Manager High Availability is configured on the cluster, login to the command line of one of the resource manager hosts, find the yarn-site.xml file (e.g. /etc/alternatives/hadoop-conf/yarn-site.xml), and from that file (similar to the below example) copy the parameters with their values into your Alpine data connection - on the Additional Parameters page (after configuring all the other settings - Integrating Spotfire Data Science with Kerberos):
yarn.resourcemanager.ha.rm-ids=rm60,rm70 yarn.resourcemanager.webapp.https.address.rm70=nn2.alpinenow.local:8090 yarn.resourcemanager.webapp.address.rm70=nn2.alpinenow.local:8088 yarn.resourcemanager.admin.address.rm70=nn2.alpinenow.local:8033 yarn.resourcemanager.resource-tracker.address.rm70=nn2.alpinenow.local:8031 yarn.resourcemanager.scheduler.address.rm70=nn2.alpinenow.local:8030 yarn.resourcemanager.address.rm70=nn2.alpinenow.local:8032 yarn.resourcemanager.webapp.https.address.rm60=nn1.alpinenow.local:8090 yarn.resourcemanager.webapp.address.rm60=nn1.alpinenow.local:8088 yarn.resourcemanager.admin.address.rm60=nn1.alpinenow.local:8033 yarn.resourcemanager.resource-tracker.address.rm60=nn1.alpinenow.local:8031 yarn.resourcemanager.scheduler.address.rm60=nn1.alpinenow.local:8030 yarn.resourcemanager.address.rm60=nn1.alpinenow.local:8032 yarn.resourcemanager.zk-address=cm.alpinenow.local:2181,nn1.alpinenow.local:2181,nn2.alpinenow.local:2181 yarn.resourcemanager.recovery.enabled=true yarn.resourcemanager.ha.automatic-failover.embedded=true yarn.resourcemanager.ha.automatic-failover.enabled=true yarn.resourcemanager.ha.enabled=true
Also, add this extra parameter with a coma-separated list of all resource manager hostnames:
failover_resource_manager_hosts=nn1.alpinenow.local,nn2.alpinenow.local
Note: If there is no TLS/SSL enabled for any of the cluster services, adding only 'failover_resource_manager_hosts' parameter should be good enough.