Cluster Management

Master/Worker Terminology is used on the Cluster UI

ACM-4086: The Master/Worker terminology naming convention is now used in the QDS UI instead of the Master/Slave terminology that was used before.

Qubole Supports New Instances

  • ACM-3436: QDS now supports c5n, m5a, and r5a instances.

Enhancements

  • ACM-3545: Qubole now provides a feature to avoid the cluster from starting whenever the master node’s Elastic IP address fails. Via Support
  • ACM-3714: While creating a Hadoop 2 (Hive) cluster, you can configure the Pig version through the Clusters UI while creating a cluster and you can also configure through the cluster API (Cluster Restart Required). Via Support
  • ACM-3848: For clusters in a VPC with multiple subnets, Qubole chooses a subnet that has sufficient IPs to launch the master node and the minimum number of nodes.
  • ACM-3833: Qubole now supports saving the node bootstrap on SSE-KMS encrypted buckets.
  • ACM-3890: You can configure 2.3 Hive (beta) version in the Configuration tab of the Clusters UI while creating a cluster. You can also configure 2.3 Hive (beta) through the cluster API.
  • ACM-3891: As part of a specific cluster’s governance, the master node is not counted in the node time chart, that is node chart only shows worker nodes.
  • ACM-4157: The Maximum Price Percentage option has replaced the Maximum Bid Price option on the Clusters UI.
  • TOOLS-978: In the AMIs, nscd is installed for DNS caching and the default value of TTL is changed to 30 seconds from 1 hour for host entries in their configuration.

Bug Fixes

  • ACM-4062: The size of log files of the httpd web server (used by Ganglia) was growing larger, which may cause disk space issues in long-running clusters. To avoid such disk-space issues, the httpd web server logs are rotated.

For a list of bug fixes between versions R54 and R55, see Changelog for api.qubole.com.