Changelog#
This file keeps track of all notable changes to the Slurm Charms.
Unreleased#
changed license to Apache-2.0
removed the action in slurmd to install InfiniBand drivers
removed the action in slurmd to install Nvidia drivers
removed the action in slurmd to install Singularity
removed the action in slurmd to install MPI (mpich)
1.0.0 - 2022-10-17#
0.10.1 - 2022-10-04#
added action in slurmd to install MPI (mpich)
0.10.0 - 2022-09-15#
updated NHC to version 1.4.3
changed how NHC is installed to be via Juju resources
added action in slurmd to install singularity
0.9.2 - 2022-07-04#
added slurmctld configuration options to use TLS certificates for etcd
0.9.1 - 2022-06-02#
added munge key to etcd database
added action in slurmctld to create etcd user to query mungekey
added action in slurmctld to get etcd root password
added action in slurmd to get etcd slurmd password
added set-node-inventory slurmd action
updated default Mellanox Infiniband drivers to version 5.4
fix get-node-inventory action
fix influxdb-info slurmctld action
fix etcd not being configured after an upgrade
0.9.0 - 2022-04-12#
fix: do not install Nvidia repository by default
add support for custom NHC parameters
improve checks when nodes need to be rebooted
add Nvidia GPU support: install drivers via an action on compute nodes
update operator framework to 1.3.0
0.8.5 - 2022-01-13#
add support for ElasticSearch Slurm addon
pin Operator Framework to 1.2.0
0.8.4 - 2021-12-10#
use the cluster name as the database name for InfluxDB
improve logrotate profiles for Slurm and NHC logs
add epilog-prolog relation
0.8.3 - 2021-11-05#
add labels to Fluentbit logs: cluster-name, partition-name, hostname, service
add user-group relation to create user and group on the system
fix typos in unit status
0.8.2 - 2021-10-13#
fix changing infiniband repos on Ubuntu
fix fluentbit parser for slurm logs
fix creation of user and group when those already exist
0.8.1 - 2021-10-07#
use Omnivector Solutions' RPM repository to install Slurm
add Fluentbit relation to forward logs from all the Slurm Charms
0.8.0 - 2021-09-13#
fix slurmd crashing when etcd in slurmctld is down/not-started/unreachable
allow installing Slurm from different repositories on Ubuntu
use Omnivector's PPA for installing Slurm on Ubuntu
0.7.0 - 2021-08-09#
fix: reduce number of events when a slurmd unit is added/removed
fix #111: install bullseye gpg keys from files
add support for Ubuntu partitions on CentOS deploys
add influxdb-info action to slurmctld
add grafana relation for slurmctld
add influxdb relation for accounting and profiling of jobs with the SLURM InfluxDB plugin
add acct gather configuration options
handle update-status hook to constantly give feedback about charms' status
add .jujuignore to reduce size of final charm files
fix cache of partition_name in slurmd charm
improve description of charm's state in juju status
fix race condition on relation data exchange between slurmctld and slurmdbd
change default partition name from juju-compute-random to osd-appname
fix slurmctld crashing when removing a slurmd application
0.6.6 - 2021-07-01#
fix checks for munge when the code can't run subprocess as another UID
0.6.5 - 2021-06-28#
fix race condition on inventory synchronization between slurmd and slurmctld before starting slurmd
fix breakage in charms when the slurmd leader is removed
remove unused code in slurmd peer relation
fix slurmd action to query infiniband version
add slurmrestd to CentOS7
fix slurmrestd Systemd environment variables to enable jwt auth
0.6.4 - 2021-06-11#
improve checks on Systemd commands to start/restart daemons, as to guarantee correct Charm initialization
check for munge keys before starting Slurm daemons, to avoid a race condition
fix possible breakage in Slurm configuration due to spaces in partition names
improve description of configurations and actions.
fix slurmrestd version in Juju status breaking the line.
0.6.3 - 2021-06-02#
fix systemd command to restart munge
improve slurmdbd logs when restarting munge
0.6.2 - 2021-06-02#
handle charm upgrade
improve juju status' message to display failure when munge/slurm does not start correctly
fix slurmd initialization when slurm.conf is present
enable munge system service, so it always start when the machine boots
0.6.1 - 2021-05-31#
changed charm's status message to have consistent capitalization.
fixed initialization order of the charms, to ensure database and controller start before compute nodes and REST server.