Base Command Manager / Bright Cluster Manager Release Notes
Release notes for Bright 7.2-33
== General ==
- New Features
* Support for magic strings in Hadoop configuration parameter values
- Fixed Issues
* Updated nag script in cm-config-rootfiles, in this version before checking the password hashes, the script will first check if the shadow file is readable.
* An issue with systemd cuda-driver.service starts too late
* JobInformation DB entries not cleaned up properly if there are too many
== cmdaemon ==
- Improvements
* Adv. conf. option NodeInfoPath to write files with category/image information for the node
* Added new mounts.pl config file to handle custom mappings for FUSE filesystems for the mounts healthcheck
- Fixed Issues
* Possible cmdaemon crash while cleaning up rsync provisioning processes
* cmdaemon not taking into account that kernel modules can be compressed
* An issue with cm-scale-cluster service in the case of slurm job arrays
* Restart of cmdaemon on the active head node results in the restart of the managed services
* Do not allow to set workload manager queue parameter in Options when the parameter is configured also via separate queue parameter (which can lead to failing slurm service)
* cmdaemon crash while running cmgui remote shell
* In HA setup, cmdaemon was executing general mirror query every 30 seconds which could cause delays in monitoring loop and could produce NULL data gaps in the monitoring data
* An issue with Slurm job queue validation in cmsh and cmgui
* cmdaemon not showing lsf/openlava job queues when the secondary head node is active
== node-installer ==
- Fixed Issues
* node-installer/busybox not taking into account that kernel modules can be compressed
* Updated boot network interface detection script and script to determine the network interface / client ID. Previously the scripts did not (correctly) detect the Mellanox ConnectX-4 Infiniband cards.
== cluster-tools ==
- Improvements
* Support for Impala on GPFS
* Support for Apache Kafka 0.10.2.0
* Support for Apache Storm 1.0.3
* Support for Cloudera CDH 5.8.4
* Support for Apache Flink 1.2.0
* Support for Cloudera CDH 5.10.0
* Support for Apache HBase 1.3.0
* Support for Cloudera CDH 5.9.1
* Support for Apache HBase 0.98.24
* Support for Apache Spark 2.1.0
* Support for Apache Flink 1.1.4
* Support for Apache Kafka 0.10.1.1
* Support for Apache Ignite 1.8.0
* Support for Apache Hive 2.1.1
* Support for Hortonworks HDP 2.5.3
* Support for Apache Drill 1.9.0
* Support for Sqoop on GPFS
* Support for Hive on GPFS
* Support for Pig on GPFS
* Support for Cloudera CDH 5.7.5
* Support for Apache Spark 2.0.2
* Support for Apache HBase 1.2.4
* Support for Cloudera CDH 5.8.3
* Support for Apache HBase 0.98.23
* Support for Cloudera CDH 5.9.0
* Support for Apache Giraph 1.2.0
* Support for Apache Kafka 0.10.1.0
* Support for Apache Zeppelin 0.6.2
* Support for Apache Flink 1.1.3
* Support for Cloudera CDH 5.7.4
* Support for Apache Spark 2.0.1
* Support for Cloudera CDH 5.8.2
- Fixed Issues
* Hortonworks HBase 2.6.x cannot start for Hadoop on GPFS
* An issue with removing the init files when disabling UGE with wlm-setup
* Replace Sqoop connector when new one is specified during re-deployment
== cm-clone-install ==
- Fixed Issues
* Slow network I/O transfer rate when booting UEFI compute nodes using TFTP
== cm-scale-cluster ==
- Improvements
* Allow for automatic removal of nodes from the sge exec hosts when the node is no longer up
- Fixed Issues
* An issue with handling UGE jobs
* cm-scale-cluster does not drain the compute node before powering off
* An issue with UGE slots calculation for cloud nodes
* In some cases, an issue with jobs being held by cmsub
== cmgui ==
- Fixed Issues
* Allow cmgui to be executed with older Firefox 56, if installed
* Hadoop setup cmgui wizard: mismatch between node selection and generated configuration for Federation scenario
== cmsh ==
- Fixed Issues
* "cmsh localhost" always redirected to the primary head node even if it is not the active, which can lead to mlnx-ofed-install.sh scripts to fail
* Cloning of a cloud node did not properly clone the EBS storage
== intel-opa ==
- Fixed Issues
* Added missing module files for MPI packages which are installed with the Intel OPA installer
== monitoring ==
- Fixed Issues
* mounts healthcheck failure for bind mounted filesystems
== node-installer-slave ==
- Fixed Issues
* Slow network I/O transfer rate when booting UEFI compute nodes using TFTP