Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 8.0-8
== General ==
- Fixed Issues
* Overlay category setting not taken into account, in pythoncm node.getAllRoles
* kubernetes: fixed an issue with KubeDNS bind address
* In some cases, an issue with monitoring producers when upgrading from 7.x to 8.0
* cmsh show in single mode does not word wrap long columns
* PDU monitoring sampler times out, and produces no data
* Clear local monitoring caches when switching to passive head node
* Status health check failed state not correct during reboot
* SGE (OGS) and UGE parallel job slots number may be calculated incorrectly by CMDaemon, cm-scale and cm-scale-cluster
* In some cases, an issue with pbs_mom directories permissions
* Power status command for azure nodes
== cmdaemon ==
- Fixed Issues
* EventLogger doesn't show Provisioning events
* Jobs with no particular nodes number request are handled incorrectly
* Improved error messages on powering on of Azure cloud nodes
* open --reset command marks the compute node as DOWN if the node was in the INSTALLING state
* Customizations should fully replace CMDaemon default values
== cluster-tools ==
- New Features
* Added new script for scheduled backups of slurm accounting database
- Improvements
* Support for Apache Hadoop 2.9.0
* Support for RDMA-Hadoop 2.x 1.3.0
* Support for Apache Hive 2.3.2
* Support for Apache ZooKeeper 3.4.11
* Support for Alluxio 1.6.1
* Support for Apache Hive 2.3.1
* Support for Apache Ignite 2.3.0
* Support for Apache Kafka 1.0.0
* Support for Hortonworks HDP 2.6.3.0
* Support for Apache Hadoop 2.8.2
- Fixed Issues
* OpenStack: race condition in HAproxy galera health check (on controller node), which could briefly disrupt connections to Galera via HAProxy
* cmhadoopctl does not parse balancer parameters
== cm-ceph-setup ==
- Fixed Issues
* Possible race condition when rebooting the OSD nodes
== cm-scale ==
- Fixed Issues
* Jobs with no particular nodes number request are handled incorrectly
* Allow to configure cm-scale in such way that it will leave unused nodes running
== hadoop ==
- Improvements
* Support for Hadoop encrypted shuffle with Security
== intel-opa ==
- Fixed Issues
* Added missing module files for MPI packages which are installed with the Intel OPA installer