Skip to content

Latest commit

 

History

History
1338 lines (1107 loc) · 55.5 KB

CHANGELOG.md

File metadata and controls

1338 lines (1107 loc) · 55.5 KB
Raw

aws-parallelcluster-cookbook CHANGELOG

This file is used to list changes made in each version of the AWS ParallelCluster cookbook.

3.12.0

ENHANCEMENTS

  • Extend Amazon DCV support to Ubuntu2204 on ARM instances.

CHANGES

  • Upgrade NVIDIA driver to version 550.127.08 (from 550.90.07). This addresses a known issue from Nivdia.
  • Upgrade Amazon DCV to version 2024.0-18131.
    • server: 2024.0-18131-1
    • xdcv: 2024.0.631-1
    • gl: 2024.0.1078-1
    • web_viewer: 2024.0-18131-1
  • Upgrade EFA installer to 1.36.0.
    • Efa-driver: efa-2.13.0-1
    • Efa-config: efa-config-1.17-1
    • Efa-profile: efa-profile-1.7-1
    • Libfabric-aws: libfabric-aws-1.22.0-1
    • Rdma-core: rdma-core-54.0-1
    • Open MPI: openmpi40-aws-4.1.7-1 and openmpi50-aws-5.0.5
  • Auto-restart slurmctld on failure.
  • Upgrade mysql-community-client to version 8.0.39.

BUG FIXES

  • Fix retrieval of regions when managing volumes to correctly handle local zones.
  • Fix an issue where adding EFS filesystems with AccessPointIds during an update would fail.

3.11.1

CHANGES

  • Pyxis is now disabled by default, so it must be manually enabled as documented in the product documentation.
  • Upgrade libjwt to version 1.17.0.

BUG FIXES

  • Fix an issue in the way we configure the Pyxis Slurm plugin in ParallelCluster that can lead to job submission failures. aws/aws-parallelcluster#6459

3.11.0

ENHANCEMENTS

  • Allow custom actions on login nodes.
  • Allow DCV connection on login nodes.
  • Add new attribute efs_access_point_ids to specify optional EFS access points for the mounts
  • Install enroot and pyxis in official pcluster AMIs

CHANGES

  • Upgrade Slurm to 23.11.10 (from 23.11.7).
  • Upgrade Pmix to 5.0.3 (from 5.0.2).
  • Upgrade EFA installer to 1.34.0.
    • Efa-driver: efa-2.10.0-1
    • Efa-config: efa-config-1.17-1
    • Efa-profile: efa-profile-1.7-1
    • Libfabric-aws: libfabric-aws-1.22.0-1
    • Rdma-core: rdma-core-52.0-1
    • Open MPI: openmpi40-aws-4.1.6-3 and openmpi50-aws-5.0.3-11
  • Upgrade NVIDIA driver to version 550.90.07 (from 535.183.01).
  • Upgrade CUDA Toolkit to version 12.4.1 (from 12.2.2).
  • Upgrade Python to 3.9.20 (from 3.9.19).
  • Upgrade Intel MPI Library to 2021.13.1.769 (from 2021.12.1.8).

BUG FIXES

  • Fix EFA kmod installation with RHEL 8.10 or newer.

3.10.1

CHANGES

  • There were no changes for this version.

3.10.0

ENHANCEMENTS

  • Add support for external Slurmdbd.
  • Add support for build-image to be run in an isolated network and ADC regions
  • Add support for Amazon Linux 2023.

CHANGES

  • Upgrade Cinc Client to version to 18.4.12 from 18.2.7.
  • Upgrade munge to version 0.5.16 (from 0.5.15).
  • Upgrade Pmix to 5.0.2 (from 4.2.9).
  • Upgrade third-party cookbook dependencies:
    • apt-7.5.22 (from apt-7.5.14)
    • openssh-2.11.12 (from openssh-2.11.3)
  • Remove third-party cookbook: selinux-6.1.12.
  • Upgrade EFA installer to 1.32.0.
    • Efa-driver: efa-2.8.0-1
    • Efa-config: efa-config-1.16-1
    • Efa-profile: efa-profile-1.7-1
    • Libfabric-aws: libfabric-aws-1.21.0-1
    • Rdma-core: rdma-core-50.0-1
    • Open MPI: openmpi40-aws-4.1.6-3 and openmpi50-aws-5.0.2-12
  • Upgrade NVIDIA driver to version 535.183.01 (from 535.154.05).
  • Upgrade Python to 3.9.19 (from 3.9.17).
  • Upgrade Intel MPI Library to 2021.12.1.8 (from 2021.9.0.43482).

BUG FIXES

  • Fix an issue that prevented cluster updates from including EFS filesystems with encryption in transit.
  • Fix an issue that prevented slurmctld and slurmdbd services from restarting on head node reboot when EFS is used for shared internal data.
  • On Ubuntu systems, remove default logrotate configuration for cloud-init log files that clashed with the configuration coming from Parallelcluster.
  • Removing /etc/profile.d/pcluster.sh so that it's not executed at every user login and cfn_bootstrap_virtualenv is not added in PATH environment variable.
  • Fix image build failure with RHEL 8.10 or newer.

3.9.3

ENHANCEMENTS

  • Add support for FSx Lustre as a shared storage type in us-iso-east-1.

BUG FIXES

  • Remove cloud_dns from the SlurmctldParameters in the Slurm config to avoid Slurm fanout issues.
    This is also not required since we set the IP addresses on instance launch.

3.9.2

CHANGES

  • Upgrade Slurm to 23.11.7 (from 23.11.4).

BUG FIXES

  • Fix issue with SharedStorageType: Efs not bootstrapping on ARM instances.

3.9.1

BUG FIXES

  • Remove recursive deletion of shared storage mountdir when unmounting filesystems as part of update-cluster operation.

3.9.0

ENHANCEMENTS

  • Permit to update the external shared storage of type Efs, FsxLustre, FsxOntap, FsxOpenZfs and FileCache without replacing compute and login fleet.
  • Add support for RHEL9.
  • Add support for Rocky Linux 9 as CustomAmi created through build-image process. No public official ParallelCluster Rocky9 Linux AMI is made available at this time.
  • Add the configuration parameter DeploymentSettings/DefaultUserHome to allow users to move the default user's home directory to /local/home instead of /home (default).
    • SSH connections will be closed and rejected while the user's home directory is being moved during the bootstrapping process.
  • Add possibility to choose between Open and Closed Source Nvidia Drivers when building an AMI, through the ['cluster']['nvidia']['kernel_open'] cookbook node attribute.
  • Add configuration parameter DeploymentSettings/DisableSudoAccessForDefaultUser to disable sudo access of default user in supported OSes.

CHANGES

  • Upgrade Slurm to 23.11.4 (from 23.02.7).
    • Upgrade Pmix to 4.2.9 (from 4.2.6).
  • Upgrade NVIDIA driver to version 535.154.05.
  • Upgrade EFA installer to 1.30.0.
    • Efa-driver: efa-2.6.0-1
    • Efa-config: efa-config-1.15-1
    • Efa-profile: efa-profile-1.6-1
    • Libfabric-aws: libfabric-aws-1.19.0
    • Rdma-core: rdma-core-46.0-1
    • Open MPI: openmpi40-aws-4.1.6-2 and openmpi50-aws-5.0.0-11
  • Upgrade NICE DCV to version 2023.1-16388.
    • server: 2023.1.16388-1
    • xdcv: 2023.1.565-1
    • gl: 2023.1.1047-1
    • web_viewer: 2023.1.16388-1
  • Upgrade ARM PL to version 23.10.
  • Upgrade third-party cookbook dependencies:
    • nfs-5.1.2 (from nfs-5.0.0)

BUG FIXES

  • Fix issue making job fail when submitted as active directory user from login nodes. The issue was caused by an incomplete configuration of the integration with the external Active Directory on the head node.
  • Fix issue making login nodes fail to bootstrap when the head node takes more time than expected in writing keys.

3.8.0

ENHANCEMENTS

  • Add support for EC2 Capacity Blocks for ML.
  • Add support for Rocky Linux 8.
  • Add support for Scheduling/SlurmSettings/Database/DatabaseName parameter to render StorageLoc in the slurmdbd configuration generated by ParallelCluster.
  • Add the option to use EFS storage instead of NFS exports from the head node root volume for intra-cluster shared file system resources: ParallelCluster, Intel, Slurm, and /home data.
  • Allow for mounting home as an EFS or FSx external shared storage via the SharedStorage section of the config file.

CHANGES

  • Upgrade Slurm to 23.02.7 (from 23.02.6).
  • Upgrade NVIDIA driver to version 535.129.03.
  • Upgrade CUDA Toolkit to version 12.2.2.
  • Use Open Source NVIDIA GPU drivers (OpenRM) as NVIDIA kernel module for Linux instead of NVIDIA closed source module.
  • Do not wait for static nodes in maintenance to signal CFN that the head node initialization is complete.
  • Upgrade EFA installer to 1.29.1.
    • Efa-driver: efa-2.6.0-1
    • Efa-config: efa-config-1.15-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.19.0-1
    • Rdma-core: rdma-core-46.0-1
    • Open MPI: openmpi40-aws-4.1.6-1
  • Upgrade GDRCopy to version 2.4 in all supported OSes, except for Centos 7 where version 2.3.1 is used.
  • Upgrade aws-cfn-bootstrap to version 2.0-28.
  • Upgrade Python to 3.9.17.

BUG FIXES

  • Fix inconsistent scaling configuration after cluster update rollback when modifying the list of instance types declared in the Compute Resources.
  • Fix users SSH keys generation when switching users without root privilege in clusters integrated with an external LDAP server through cluster configuration files.
  • Fix disabling Slurm power save mode when setting ScaledownIdletime = -1.
  • Fix hard-coded path to Slurm installation dir in update_slurm_database_password.sh script for Slurm Accounting.

3.7.2

CHANGES

  • Upgrade Slurm to 23.02.6.

3.7.1

CHANGES

  • Upgrade Slurm to 23.02.5 (from 23.02.4).
    • Upgrade Pmix to 4.2.6 (from 3.2.3).
    • Upgrade libjwt to 1.15.3 (from 1.12.0).
  • Upgrade EFA installer to 1.26.1, fixing RDMA writedata issue in P5.
    • Efa-driver: efa-2.5.0-1
    • Efa-config: efa-config-1.15-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.18.2-1
    • Rdma-core: rdma-core-46.0-1
    • Open MPI: openmpi40-aws-4.1.5-4

3.7.0

ENHANCEMENTS

  • Allow configuration of static and dynamic node priorities in compute resources via the ParallelCluster configuration YAML file.
  • Forward SLURM_RESUME_FILE to ParallelCluster resume program.
  • Allow to override aws-parallelcluster-node package at cluster creation and update time (only on the head node during update). Useful for development purposes only.
  • Avoid to start NFS server on compute nodes.
  • Add support for Login Nodes.

CHANGES

  • Assign Slurm dynamic nodes a priority (weight) of 1000 by default. This allows Slurm to prioritize idle static nodes over idle dynamic ones.
  • Create a Slurm partition-nodelist mapping JSON file to be used by the node package daemons to recognize PC-managed Slurm partitions and nodelists.
  • Upgrade NVIDIA driver to version 535.54.03.
  • Upgrade CUDA library to version 12.2.0.
  • Upgrade NVIDIA Fabric manager to nvidia-fabricmanager-535
  • Increase EFS-utils watchdog poll interval to 10 seconds. Note: This change is meaningful only if EncryptionInTransit is set to true, because watchdog does not run otherwise.
  • Upgrade EFA installer to 1.25.1
    • Efa-driver: efa-2.5.0-1
    • Efa-config: efa-config-1.15-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.18.1-1
    • Rdma-core: rdma-core-46.0-1
    • Open MPI: openmpi40-aws-4.1.5-4
  • Upgrade Slurm to version 23.02.4.
  • Upgrade ARM PL to version 23.04.1 for Ubuntu 22.04 only.
  • Upgrade third-party cookbook dependencies:
    • apt-7.5.14 (from apt-7.4.0)
    • line-4.5.13 (from line-4.5.2)
    • openssh-2.11.3 (from openssh-2.10.3)
    • pyenv-4.2.3 (from pyenv-3.5.1)
    • selinux-6.1.12 (from selinux-6.0.5)
    • yum-7.4.13 (from yum-7.4.0)
    • yum-epel-5.0.2 (from yum-epel-4.5.0)
  • Upgrade NICE DCV to version 2023.0-15487.
    • server: 2023.0.15487-1
    • xdcv: 2023.0.551-1
    • gl: 2023.0.1039-1
    • web_viewer: 2023.0.15487-1

BUG FIXES

  • Fix cluster creation failure with Ubuntu Deep Learning AMI on GPU instances and DCV enabled.

3.6.1

CHANGES

  • Remove security updates step executed on cluster nodes bootstrap in US isolated regions in order to reduce bootstrap time and avoid a potential point of failure.
  • Replace nvidia-persistenced service with parallelcluster_nvidia service to avoid conflicts with DLAMI.

BUG FIXES

  • Fix an issue that was preventing ptrace protection from being disabled on Ubuntu and allowing Cross Memory Attach (CMA) in libfabric.

3.6.0

ENHANCEMENTS

  • Add support for RHEL8.
  • Add support for customizing the cluster Slurm configuration via the ParallelCluster configuration YAML file.
  • Build Slurm with support for LUA.
  • Add health check manager and GPU health check, which can be activated through cluster configuration. Health check manager execution is triggered by a Slurm prolog script. GPU check verifies healthiness of a node by executing NVIDIA DCGM L2 diagnostic.
  • Add log rotation support for ParallelCluster managed logs.
  • Track head node memory and root volume disk utilization using the mem_used_percent and disk_used_percent metrics collected through the CloudWatch Agent.
  • Enforce the DCV Authenticator Server to use at least TLS-1.2 protocol when creating the SSL Socket.
  • Load kernel module nvidia-uvm by default to provide Unified Virtual Memory (UVM) functionality to the CUDA driver.
  • Install NVIDIA Persistence Daemon as a system service.
  • Install NVIDIA Data Center GPU Manager (DCGM) package on all supported OSes except for aarch64 centos7 and alinux2.

CHANGES

  • Upgrade Slurm to version 23.02.2.
  • Upgrade munge to version 0.5.15.
  • Set Slurm default TreeWidth to 30.
  • Set Slurm prolog and epilog configurations to target a directory, /opt/slurm/etc/scripts/prolog.d/ and /opt/slurm/etc/scripts/epilog.d/ respectively.
  • Set Slurm BatchStartTimeout to 3 minutes so to allow max 3 minutes Prolog execution during compute node registration.
  • Upgrade EFA installer to 1.22.1
    • Dkms : 2.8.3-2
    • Efa-driver: efa-2.1.1g
    • Efa-config: efa-config-1.13-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.17.1-1
    • Rdma-core: rdma-core-43.0-1
    • Open MPI: openmpi40-aws-4.1.5-1
  • Upgrade Lustre client version to 2.12 on Amazon Linux 2 (same version available on Ubuntu 20.04, 18.04 and CentOS >= 7.7).
  • Upgrade Lustre client version to 2.10.8 on CentOS 7.6.
  • Upgrade aws-cfn-bootstrap to version 2.0-24.
  • Upgrade NVIDIA driver to version 470.182.03.
  • Upgrade NVIDIA Fabric Manager to version 470.182.03.
  • Upgrade NVIDIA CUDA Toolkit to version 11.8.0.
  • Upgrade NVIDIA CUDA sample to version 11.8.0.
  • Upgrade Intel MPI Library to 2021.9.0.43482.
  • Upgrade NICE DCV to version 2023.0-15022.
    • server: 2023.0.15022-1
    • xdcv: 2023.0.547-1
    • gl: 2023.0.1027-1
    • web_viewer: 2023.0.15022-1
  • Avoid to reset FSx and EFS shared folders permissions when mounting them in the compute nodes.

BUG FIXES

  • Fix an issue that was causing misalignment of compute nodes IP on instances with multiple network interfaces.
  • Fix replacement of StoragePass in slurm_parallelcluster_slurmdbd.conf when a queue parameter update is performed and the Slurm accounting configurations are not updated.
  • Fix issue causing cfn-hup daemon to fail when it gets restarted.
  • Fix issue causing NVIDIA GPU compute nodes not to resume correctly after executing an scontrol reboot command.

3.5.1

ENHANCEMENTS

  • Add support for US isolated region us-isob-east-1.

CHANGES

  • Upgrade EFA installer to 1.22.0
    • Efa-driver: efa-2.1.1g
    • Efa-config: efa-config-1.13-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.17.0-1
    • Rdma-core: rdma-core-43.0-1
    • Open MPI: openmpi40-aws-4.1.5-1
  • Upgrade NICE DCV to version 2022.2-14521.
    • server: 2022.2.14521-1
    • xdcv: 2022.2.519-1
    • gl: 2022.2.1012-1
    • web_viewer: 2022.2.14521-1

BUG FIXES

  • Fix update cluster to remove shared EBS volumes can potentially cause node launching failures if MountDir match the same pattern in /etc/exports.

3.5.0

ENHANCEMENTS

  • Fail cluster creation if cluster status changes to PROTECTED while provisioning static nodes.

CHANGES

  • Upgrade Slurm to version 22.05.8.
  • Upgrade EFA installer to 1.21.0
    • Efa-driver: efa-2.1.1-1
    • Efa-config: efa-config-1.12-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.16.1amzn3.0-1
    • Rdma-core: rdma-core-43.0-1
    • Open MPI: openmpi40-aws-4.1.4-3
  • Make Slurm controller logs more verbose and enable additional logging for the Slurm power save plugin.

BUG FIXES

  • Fix an issue where custom AMI creation failed in Ubuntu 20.04 on MySQL packages installation.

3.4.1

BUG FIXES

  • Fix an issue with the Slurm scheduler that might incorrectly apply updates to its internal registry of compute nodes. This might result in EC2 instances to become inaccessible or backed by an incorrect instance type.

3.4.0

ENHANCEMENTS

  • Add support for specifying multiple subnets for each queue to increase the EC2 capacity pool available for use.

CHANGES

  • Upgrade EFA installer to 1.20.0
    • Efa-driver: efa-2.1
    • Efa-config: efa-config-1.11-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.16.1
    • Rdma-core: rdma-core-43.0-2
    • Open MPI: openmpi40-aws-4.1.4-3
  • Mount EFS file systems using amazon-efs-utils. EFS files systems can be mounted using in-transit encryption and IAM authorized user.
  • Install stunnel 5.67 on CentOS7 and Ubuntu to support EFS in-transit encryption.
  • Add possibility to execute a custom script in the head node during the update of the cluster.
  • Upgrade Slurm to version 22.05.7.
  • Upgrade Python to 3.9.16 and 3.7.16.

3.3.0

ENHANCEMENTS

  • Add support for Slurm Accounting.
  • Add support for adding and removing shared storages at cluster update.
  • Add possibility to specify multiple instance types for the same compute resource.
  • Configure NFS threads to be min(256, max(8, num_cores * 4)) to ensure better stability and performance.
  • Move NFS installation at build time to reduce configuration time.

CHANGES

  • Upgrade NVIDIA driver to version 470.141.03.
  • Upgrade NVIDIA Fabric Manager to version 470.141.03.
  • Upgrade NVIDIA CUDA Toolkit to version 11.7.1.
  • Disable cron job tasks man-db and mlocate, which may have a negative impact on node performance.
  • Reduce timeout from 50 to a maximum of 5min in case of DynamoDB connection issues at compute node bootstrap.
  • Change the logic to number the routing tables when an instance have multiple NICs.
  • Upgrade Python from 3.7.13 to 3.9.15.
  • Upgrade Slurm to version 22.05.5.
  • Upgrade EFA installer to 1.18.0.
    • Efa-driver: efa-1.16.0-1
    • Efa-config: efa-config-1.11-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.16.0~amzn4.0-1
    • Rdma-core: rdma-core-41.0-2
    • Open MPI: openmpi40-aws-4.1.4-2
  • Upgrade NICE DCV to version 2022.1-13300.
    • server: 2022.1.13300-1
    • xdcv: 2022.1.433-1
    • gl: 2022.1.973-1
    • web_viewer: 2022.1.13300-1
  • Upgrade third-party cookbook dependencies:
    • selinux-6.0.5 (from selinux-6.0.4)
    • nfs-5.0.0 (from nfs-2.6.4)

3.2.1

ENHANCEMENTS

  • Improve the logic to associate the host routing tables to the different network cards to better support EC2 instances with several NICs.

CHANGES

  • Upgrade NVIDIA driver to version 470.141.03.
  • Upgrade NVIDIA Fabric Manager to version 470.141.03.
  • Pin cfn-bootstrap helper package version to 2.0-10
  • Disable cron job tasks man-db and mlocate, which may have a negative impact on node performance.
  • Upgrade Intel MPI Library to 2021.6.0.602.
  • Upgrade Python from 3.7.10 to 3.7.13 in response to this security risk.

3.2.0

ENHANCEMENTS

  • Add support for multiple Elastic File Systems.
  • Add support for multiple FSx File System.
  • Add support for attaching existing FSx for Ontap and FSx for OpenZFS File Systems.
  • Install NVIDIA GDRCopy 2.3 to enable low-latency GPU memory copy on supported instance types.
  • During cluster update set Slurm nodes state accordingly to strategy set through the configuration parameter Scheduling/SchedulerSettings/QueueUpdateStrategy.
  • Add support for memory-based scheduling in Slurm.
    • Configure RealMemory on compute nodes by default as 95% of the EC2 memory.
    • Move SelectTypeParameters to slurm_parallelcluster.conf include file.
    • Move ConstrainRAMSpace to slurm_parallelcluster_cgroup.conf include file.
    • Add support for new configuration parameter Scheduling/SlurmSettings/EnableMemoryBasedScheduling to configure memory-based scheduling in Slurm.
    • Add support for new configuration parameter Scheduling/SlurmQueues/ComputeResources/SchedulableMemory to override default value of the memory seen by the scheduler on compute nodes.
  • Add support for rebooting compute nodes via Slurm.

CHANGES

  • Restart clustermgtd and slurmctld daemons at cluster update time only when Scheduling parameters are updated in the cluster configuration.
  • Update slurmctld and slurmd systemd service files.
  • Upgrade NICE DCV to version 2022.0-12760.
  • Upgrade NVIDIA driver to version 470.129.06.
  • Upgrade NVIDIA Fabric Manager to version 470.129.06.
  • Upgrade EFA installer to version 1.17.2.
    • EFA driver: efa-1.16.0-1
    • EFA configuration: efa-config-1.10-1
    • EFA profile: efa-profile-1.5-1
    • Libfabric: libfabric-aws-1.16.0~amzn2.0-1
    • RDMA core: rdma-core-41.0-2
    • Open MPI: openmpi40-aws-4.1.4-2
  • Restrict IPv6 access to IMDS to root and cluster admin users only, when configuration parameter HeadNode/Imds/Secured is enabled.
  • Set Slurm configuration AuthInfo=cred_expire=70 to reduce the time requeued jobs must wait before starting again when nodes are not available.
  • Move SelectTypeParameters and ConstrainRAMSpace to the parallelcluster_slurm*.conf include files.
  • Upgrade third-party cookbook dependencies:
    • apt-7.4.2 (from apt-7.4.0)
    • line-4.5.13 (from line-4.0.1)
    • openssh-2.10.3 (from openssh-2.9.1)
    • pyenv-3.5.1 (from pyenv-3.4.2)
    • selinux-6.0.4 (from selinux-3.1.1)
    • yum-7.4.0 (from yum-6.1.1)
    • yum-epel-4.5.0 (from yum-epel-4.1.2)
  • Disable aws-ubuntu-eni-helper service, available in Deep Learning AMIs, to avoid conflicts with configure_nw_interface.sh when configuring instances with multiple network cards.
  • Set MTU to 9001 for all the network interfaces when configuring instances with multiple network cards.
  • Remove the trailing dot when configuring the compute node FQDN.

3.1.5

CHANGES

  • Upgrade EFA installer to 1.18.0
    • Efa-driver: efa-1.16.0-1
    • Efa-config: efa-config-1.11-1
    • Efa-profile: efa-profile-1.5-1
    • Libfabric-aws: libfabric-aws-1.16.0~amzn4.0-1
    • Rdma-core: rdma-core-41.0-2
    • Open MPI: openmpi40-aws-4.1.4-2
  • Upgrade Intel MPI Library to 2021.6.0.602.
  • Upgrade NVIDIA driver to version 470.141.03.
  • Upgrade NVIDIA Fabric Manager to version 470.141.03.

BUG FIXES

  • Fix Slurm issue that prevents idle nodes termination.

3.1.4

CHANGES

  • Upgrade Slurm to version 21.08.8.

ENHANCEMENTS

  • Add support for enabling JWT authentication Slurm.

3.1.3

ENHANCEMENTS

  • Execute SSH key creation alongside with the creation of HOME directory, i.e. during SSH login, when switching to another user and when executing a command as another user.
  • Add support for both FQDN and LDAP Distinguished Names in the configuration parameter DirectoryService/DomainName. The new validator now checks both the syntaxes.
  • New update_directory_service_password.sh script deployed on the head node supports the manual update of the Active Directory password in the SSSD configuration. The password is retrieved by the AWS Secrets Manager as from the cluster configuration.
  • Make DirectoryService/AdditionalSssdConfigs be merged into final SSSD configuration rather than be appended.

CHANGES

  • Disable deeper C-States in x86_64 official AMIs and AMIs created through build-image command, to guarantee high performance and low latency.
  • Change Amazon Linux 2 base images to use AMIs with Kernel 5.10.

BUG FIXES

  • Fix the configuration parameter DirectoryService/DomainAddr conversion to ldap_uri SSSD property when it contains multiples domain addresses.
  • Fix DCV not loading user profile at session start. The user's PATH was not correctly set at DCV session connection.

3.1.2

CHANGES

  • Upgrade Slurm to version 21.08.6.

BUG FIXES

  • Fix the update of /etc/hosts file on computes nodes when a cluster is deployed in subnets without internet access.
  • Fix compute nodes bootstrap by waiting for ephemeral drives initialization before joining the cluster.

3.1.1

ENHANCEMENTS

  • Add support for multiple users cluster environments by integrating with Active Directory (AD) domains managed via AWS Directory Service.
  • Install NVIDIA drivers and CUDA library for ARM.
  • Build Slurm with slurmrestd support.

CHANGES

  • Upgrade Slurm to version 21.08.5.
  • Upgrade NICE DCV to version 2021.3-11591.
  • Upgrade NVIDIA driver to version 470.103.01.
  • Upgrade CUDA library to version 11.4.4.
  • Upgrade NVIDIA Fabric manager to version 470.103.01.
  • Upgrade Intel MPI Library to 2021.4.0.441.
  • Upgrade PMIx to version 3.2.3.
  • Move the configure/install recipes to separate cookbooks that are called from the main one. Existing entrypoints are maintained and backwards compatible.
  • Download dependencies of Intel HPC platform during AMI build time to avoid contacting internet during cluster creation time.
  • Do not strip - from compute resource name when configuring Slurm nodes.
  • Add cluster parameter directory_service.disabled_on_compute_nodes to disable AD integration on compute nodes.

BUG FIXES

  • Do not configure GPUs in Slurm when NVIDIA driver is not installed.
  • Fix the way ExtraChefAttributes are merged into the final configuration.
  • Fix sporadic failure of build-image due to missing package

3.0.3

CHANGES

  • Disable log4j-cve-2021-44228-hotpatch service on Amazon Linux to avoid incurring in potential performance degradation.

3.0.2

CHANGES

  • Upgrade EFA installer to version 1.14.1. Thereafter, EFA enables GDR support by default on supported instance type(s). ParallelCluster does not reinstall EFA during node start. Previously, EFA was reinstalled if GdrSupport had been turned on in the configuration file.
    • EFA configuration: efa-config-1.9-1
    • EFA profile: efa-profile-1.5-1
    • EFA kernel module: efa-1.14.2
    • RDMA core: rdma-core-37.0
    • Libfabric: libfabric-1.13.2
    • Open MPI: openmpi40-aws-4.1.1-2

BUG FIXES

  • Fix issue that is preventing cluster names to start with parallelcluster- prefix.

3.0.1

CHANGES

  • Change supervisord service script from SysVinit to Systemd.
  • Drop support for SysVinit. Only Systemd is supported.

BUG FIXES

  • Fix supervisord service not enabled on Ubuntu.
  • Update ca-certificates package during AMI build time and prevent Chef from using outdated/distrusted CA certificates.

3.0.0

ENHANCEMENTS

  • Support restart/reboot for instance type with instance store (ephemeral drives).
  • Compile Slurm with jobcomp/elasticsearch support.

CHANGES

  • Update pre / post install scripts to drop the first argument which was the script being run
  • Drop support for SGE and Torque schedulers.
  • Drop support for CentOS8.
  • Remove nodewatcher, sqswatcher, jobwatcher related code.
  • Remove Ganglia support.
  • Install ParallelCluster AWS Batch CLI at AMI build time.
  • Run daemons as cluster admin user (not root).
  • Add explicit assignment of names, uids, gids for slurm, munge and dcvextauth users.
  • Remove packer.
  • Restrict access to IMDS to root and cluster admin users, only.
  • Make PATH include required directories for every user and recipes context.
  • Fail cluster creation when IMDS lockdown is not working correctly.
  • Make sudoers secure_path include the same directories in every platform.
  • Remove option for instance store software encryption (encrypted_ephemeral).
  • Add support for iptables restore on instance reboot.
  • Allow IMDS access for dcv user when dcv is enabled.
  • Restore noatime option, which has positive impact on the performances of NFS filesystem
  • Upgrade NICE DCV to version 2021.1-10851.
  • Upgrade Slurm to version 20.11.8
  • Upgrade Cinc Client to version 17.2.29.
  • Upgrade NVIDIA driver to version 470.57.02.
  • Upgrade CUDA library to version 11.4.0.
  • Avoid installing MPICH and FFTW packages.
  • Upgrade EFA installer to version 1.13.0
    • Update rdma-core to v35.0.
    • Update libfabric to v1.13.0amzn1.0.

BUG FIXES

  • Fix cluster update when using proxy setup

2.11.3

CHANGES

  • Upgrade EFA installer to version 1.14.1. Thereafter, EFA enables GDR support by default on supported instance type(s). ParallelCluster does not reinstall EFA during node start. Previously, EFA was reinstalled if enable_efa_gdr had been turned on in the configuration file.
    • EFA configuration: efa-config-1.9-1
    • EFA profile: efa-profile-1.5-1
    • EFA kernel module: efa-1.14.2
    • RDMA core: rdma-core-37.0
    • Libfabric: libfabric-1.13.2
    • Open MPI: openmpi40-aws-4.1.1-2

BUG FIXES

  • Fix failure when building AMI, due to SGE sources not available at arc.liv.ac.uk
  • Fix cluster update when using proxy setup.
  • Update ca-certificates package during AMI build time and prevent Chef from using outdated/distrusted CA certificates.

2.11.2

CHANGES When using a custom AMI with a preinstalled EFA package, no actions are taken at node bootstrap time in case GPUDirect RDMA is enabled. The original EFA package deployment is preserved as during the Image build process.

BUG FIXES

  • Lock version of nvidia-fabricmanager package to prevent updates and misalignments with NVIDIA drivers

2.11.1

ENHANCEMENTS

  • Retry failed installations of aws-parallelcluster package on head node of clusters using AWS Batch as the scheduler.

CHANGES

  • Restore noatime option, which has positive impact on the performances of NFS filesystem.
  • Upgrade EFA installer to version 1.12.3
    • EFA configuration: efa-config-1.9 (from efa-config-1.8-1)
    • EFA kernel module: efa-1.13.0 (from efa-1.12.3)

BUG FIXES

  • Pin to version 1.247347 of the CloudWatch agent due to performance impact of latest CW agent version 1.247348.
  • Avoid failures when building SGE using instance type with vCPU >=32.

2.11.0

ENHANCEMENTS

  • Add support for Ubuntu 20.04.
  • Add support for using FSx Lustre in subnet with no internet access.
  • Add support for building Centos 7 AMIs on ARM.
  • Make sure slurmd service is only enabled after post-install process, which will prevent user from unintentionally making compute node available during post-install process.
  • Change to ssh_target_checker.sh syntax that makes the script compatible with pdsh.
  • Add possibility to use a post installation script when building Centos 8 AMI.
  • Install SSM agent on CentOS 7 and 8.
  • Transition from IMDSv1 to IMDSv2.
  • Add support for security_group_id in packer custom builders. Customers can export AWS_SECURITY_GROUP_ID environment variable to specify security group for custom builders when building custom AMIs.
  • Configure the following default gc_thresh values for performance at scale.
    • net.ipv4.neigh.default.gc_thresh1 = 0
    • net.ipv4.neigh.default.gc_thresh2 = 15360
    • net.ipv4.neigh.default.gc_thresh3 = 16384

CHANGES

  • Ubuntu 16.04 is no longer supported.
  • Amazon Linux is no longer supported.
  • Upgrade EFA installer to version 1.12.2
    • EFA configuration: efa-config-1.8-1 (from efa-config-1.7)
    • EFA profile: efa-profile-1.5-1 (from efa-profile-1.4)
    • EFA kernel module: efa-1.12.3 (from efa-1.10.2)
    • RDMA core: rdma-core-32.1amzn (from rdma-core-31.2amzn)
    • Libfabric: libfabric-1.11.2amzon1.1-1 (from libfabric-1.11.1amzn1.0)
    • Open MPI: openmpi40-aws-4.1.1-2 (from openmpi40-aws-4.1.0)
  • Increase timeout when attaching EBS volumes from 3 to 5 minutes.
  • Retry berkshelf installation up to 3 times.
  • Root volume size increased from 25GB to 35GB on all AMIs. Minimum root volume size is now 35GB.
  • Upgrade Slurm to version 20.11.7.
    • Update slurmctld and slurmd systemd unit files according to latest provided by slurm
    • Add new SlurmctldParameters, power_save_min_interval=30, so power actions will be processed every 30 seconds
    • Specify instance GPU model as GRES GPU Type in gres.conf, instead of previous hardcoded value Type=tesla for all GPU
  • Upgrade Arm Performance Libraries (APL) to version 21.0.0
  • Upgrade NICE DCV to version 2021.1-10557.
  • Upgrade NVIDIA driver to version 460.73.01.
  • Upgrade CUDA library to version 11.3.0.
  • Upgrade NVIDIA Fabric manager to nvidia-fabricmanager-460.
  • Install ParallelCluster AWSBatch CLI in dedicated python3 virtual env.
  • Upgrade Python version used in ParallelCluster virtualenvs from version 3.6.13 to version 3.7.10.
  • Upgrade Cinc Client to version 16.13.16.
  • Upgrade third-party cookbook dependencies:
    • apt-7.4.0 (from apt-7.3.0)
    • iptables-8.0.0 (from iptables-7.1.0)
    • line-4.0.1 (from line-2.9.0)
    • openssh-2.9.1 (from openssh-2.8.1)
    • pyenv-3.4.2 (from pyenv-3.1.1)
    • selinux-3.1.1 (from selinux-2.1.1)
    • ulimit-1.1.1 (from ulimit-1.0.0)
    • yum-6.1.1 (from yum-5.1.0)
    • yum-epel-4.1.2 (from yum-epel-3.3.0)
  • Drop lightdm package install from Ubuntu 18.04 DCV installation process.
  • Update default NFS options used by Compute nodes to mount shared filesystem from head node.
    • Drop intr option, which is deprecated since kernel 2.6.25
    • Drop noatime option, which is not relevant for NFS mount

2.10.4

CHANGES

  • Upgrade Slurm to version 20.02.7

2.10.3

CHANGES

  • Upgrade EFA installer to version 1.11.2
    • EFA configuration: efa-config-1.7 (no change)
    • EFA profile: efa-profile-1.4 (from efa-profile-1.3)
    • EFA kernel module: efa-1.10.2 (no change)
    • RDMA core: rdma-core-31.2amzn (no change)
    • Libfabric: libfabric-1.11.1amzn1.0 (no change)
    • Open MPI: openmpi40-aws-4.1.0 (no change)

2.10.2

ENHANCEMENTS

  • Improve configuration procedure for the Munge service.

CHANGES

  • Update Python version used in ParallelCluster virtualenvs from version 3.6.9 to version 3.6.13.

BUG FIXES

  • Use non interactive apt update command when building custom Ubuntu AMIs.
  • Fix encrypted_ephemeral = true when using Alinux2 or CentOS8

2.10.1

ENHANCEMENTS

  • Install Arm Performance Libraries (APL) 20.2.1 on ARM AMIs (CentOS8, Alinux2, Ubuntu1804).
  • Install EFA kernel module on ARM instances with alinux2 and ubuntu1804.
  • Configure NFS threads to be max(8, num_cores) for performance. This enhancement will not take effect on Ubuntu 16.04.

CHANGES

  • Upgrade EFA installer to version 1.11.1.
    • EFA configuration: efa-config-1.7 (from efa-config-1.5)
    • EFA profile: efa-profile-1.3 (from efa-profile-1.1)
    • EFA kernel module: efa-1.10.2 (no change)
    • RDMA core: rdma-core-31.2amzn (from rdma-core-31.amzn0)
    • Libfabric: libfabric-1.11.1amzn1.0 (from libfabric-1.11.1amzn1.1)
    • Open MPI: openmpi40-aws-4.1.0 (from openmpi40-aws-4.0.5)
  • Upgrade Intel MPI to version U8.
  • Upgrade NICE DCV to version 2020.2-9662.
  • Set default systemd runlevel to multi-user.target on all OSes during ParallelCluster official ami creation. The runlevel is set to graphical.target on head node only when DCV is enabled. This prevents the execution of graphical services, such as x/gdm, when they are not required.
  • Download Intel MPI and HPC packages from S3 rather than Intel yum repos.

BUG FIXES

  • Fix installation of Intel PSXE package on CentOS 7 by using yum4.
  • Fix routing issues with multiple Network Interfaces on Ubuntu 18.04.
  • Fix compilation of SGE by downloading sources from Debian repository and not from the EOL Ubuntu 19.10.

2.10.0

ENHANCEMENTS

  • Add support for CentOS 8.
  • Add support for instance types with multiple network cards (e.g. p4d.24xlarge).
  • Enable FSx Lustre in China regions.
  • Add validation step for AMI creation process to fail when using a base AMI created by a different version of ParallelCluster.
  • Add validation step for AMI creation process to fail if the selected OS and the base AMI OS are not consistent.
  • Add possibility to use a post installation script when building an AMI.
  • Install NVIDIA Fabric manager to enable NVIDIA NVSwitch on supported platforms.

CHANGES

  • Upgrade EFA installer to version 1.10.1
    • EFA configuration: efa-config-1.5 (from efa-config-1.4)
    • EFA profile: efa-profile-1.1 (from efa-profile-1.0.0)
    • EFA kernel module: efa-1.10.2 (from efa-1.6.0)
    • RDMA core: rdma-core-31.amzn0 (from rdma-core-28.amzn0)
    • Libfabric: libfabric-1.11.1amzn1.1 (from libfabric-1.10.1amzn1.1)
    • Open MPI: openmpi40-aws-4.0.5 (from openmpi40-aws-4.0.3)
    • Unifies installer runtime options across x86 and aarch64
    • Introduces -g/--enable-gdr switch to install packages with GPUDirect RDMA support
    • Updates to OMPI collectives decision file packaging, migrated from efa-config to efa-profile
    • Introduces CentOS 8 support
  • CentOS 6 is no longer supported.
  • Upgrade NVIDIA driver to version 450.80.02.
  • Upgrade Intel Parallel Studio XE Runtime to version 2020.2.
  • Upgrade Munge to version 0.5.14.
  • Retrieve FSx Lustre DNS name dynamically.
  • Slurm: change SlurmctldPort to 6820-6829 to not overlap with default slurmdbd port (6819).
  • Slurm: add compute_resource name and efa as node features.
  • Improve Slurm and Munge installation process by cleaning up existing installations from OS repositories.
  • Install Python 3 version of aws-cfn-bootstrap scripts.
  • Do not force compute fleet into STOPPED state when performing a cluster update. This allows to update the queue size without forcing a termination of the existing instances.

BUG FIXES

  • Fix ephemeral drives setup to avoid failures when partition changes require a reboot.
  • Fix Chrony service management.
  • Retrieve the right number of compute instance slots when instance type is updated.
  • Fix compute fleet status initialization to be configured before daemons are started by supervisord.

2.9.1

CHANGES

  • There were no notable changes for this version.

2.9.0

ENHANCEMENTS

  • Add support for multiple queues and multiple instance types feature with the Slurm scheduler.
  • Extend NICE DCV support to ARM instances.
  • Extend support to disable hyperthreading on instances (like *.metal) that don't support CpuOptions in LaunchTemplate.
  • Enable support for NFS 4 for the filesystems shared from the head node.
  • Add script wrapper to support Torque-like commands with the Slurm scheduler.

CHANGES

  • A Route53 private hosted zone is now created together with the cluster and used in DNS resolution inside cluster nodes when using Slurm scheduler.
  • Upgrade EFA installer to version 1.9.5:
    • EFA configuration: efa-config-1.4 (from efa-config-1.3)
    • EFA profile: efa-profile-1.0.0
    • EFA kernel module: efa-1.6.0 (no change)
    • RDMA core: rdma-core-28.amzn0 (no change)
    • Libfabric: libfabric-1.10.1amzn1.1 (no change)
    • Open MPI: openmpi40-aws-4.0.3 (no change)
  • Upgrade Slurm to version 20.02.4.
  • Apply the following changes to Slurm configuration:
    • Assign a range of 10 ports to Slurmctld in order to better perform with large cluster settings
    • Configure cloud scheduling logic
    • Set ReconfigFlags=KeepPartState
    • Set MessageTimeout=60
    • Set TaskPlugin=task/affinity,task/cgroup together with TaskAffinity=no and ConstrainCores=yes in cgroup.conf
  • Upgrade NICE DCV to version 2020.1-9012.
  • Use private ip instead of master node hostname when mounting shared NFS drives.
  • Add new log streams to CloudWatch: chef-client, clustermgtd, computemgtd, slurm_resume, slurm_suspend.
  • Remove dependency on cfn-init in compute nodes bootstrap.
  • Add support for queue names in pre/post install scripts.

BUG FIXES

  • Solve dpkg lock issue with Ubuntu that prevented custom AMI creation in some cases.

2.8.1

CHANGES

  • Disable screen lock for DCV desktop sessions to prevent users from being locked out.

2.8.0

ENHANCEMENTS

  • Enable support for ARM instances on Ubuntu 18.04 and Amazon Linux 2.
  • Install PMIx v3.1.5 and provide slurm support for it on all supported operating systems except for CentOS 6.
  • Install glibc-static, which is required to support certain options for the Intel MPI compiler.

CHANGES

  • Disable libvirtd service on Centos 7. Virtual bridge interfaces are incorrectly detected by Open MPI and cause MPI applications to hang, see https://www.open-mpi.org/faq/?category=tcp#tcp-selection for details
  • Use CINC instead of Chef for provisioning instances. See https://cinc.sh/about/ for details.
  • Retry when mounting an NFS mount fails.
  • Install the pyenv virtual environments used by ParallelCluster cookbook and node daemon code under /opt/parallelcluster instead of under /usr/local.
  • Avoid downloading the source for env2 at installation time.
  • Drop dependency on the gems ridley and ffi-libarchive.
  • Vendor cookbooks as part of instance provisioning, rather than doing so before copying the cookbook into an instance. Users no longer need to have berks installed locally.
  • Drop the dependencies on the poise-python, tar and hostname third-party cookbooks.
  • Use the new official CentOS 7 AMI as the base images for ParallelCluster AMI.
  • Upgrade NVIDIA driver to Tesla version 440.95.01 on CentOS 6 and version 450.51.05 on all other distros.
  • Upgrade CUDA library to version 11.0 on all distros besides CentOS 6.
  • Install third-party cookbook dependencies via local source, rather than using the Chef supermarket.
  • Use https wherever possible in download URLs.
  • Upgrade EFA installer to version 1.9.4:
    • Kernel module: efa-1.6.0 (from efa-1.5.1)
    • RDMA core: rdma-core-28.amzn0 (from rdma-core-25.0)
    • Libfabric: libfabric-1.10.1amzn1.1 (updated from libfabric-aws-1.9.0amzn1.1)
    • Open MPI: openmpi40-aws-4.0.3 (no change)

BUG FIXES

  • Fix issue that was preventing concurrent use of custom node and pcluster CLI packages.
  • Use the correct domain name when contacting AWS services from the China partition.
  • Avoid pinning to a specific release of the Intel HPC platform.

2.7.0

CHANGES

  • Upgrade NICE DCV to version 2020.0-8428.
  • Upgrade Intel MPI to version U7.
  • Upgrade NVIDIA driver to version 440.64.00.
  • Upgrade EFA installer to version 1.8.4:
    • Kernel module: efa-1.5.1 (no change)
    • RDMA core: rdma-core-25.0 (no change)
    • Libfabric: libfabric-aws-1.9.0amzn1.1 (no change)
    • Open MPI: openmpi40-aws-4.0.3 (updated from openmpi40-aws-4.0.2)
  • Upgrade CentOS 7 AMI to version 7.8

BUG FIXES

  • Fix recipes installation at runtime by adding the bootstrapped file at the end of the last chef run.
  • Fix installation of Lustre client on Centos 7
  • FSx Lustre: Exit with error when failing to retrieve FSx mountpoint.

2.6.2

ENHANCEMENTS

CHANGES

  • Upgrade Intel MPI to version 2019 U7.

  • Upgrade EFA installer to version 1.8.4

    • Kernel module: efa-1.5.1 (no change)
    • RDMA core: rdma-core-25.0 (distributed only) (no change)
    • Libfabric: libfabric-aws-1.9.0amzn1.1 (no change)
    • Open MPI: openmpi40-aws-4.0.3 (updated from openmpi40-aws-4.0.2)
    • EFA profile: efa-profile-1.0 (new)
      • Configures $PATH and ld config bits for Libfabric and Open MPI installed by EFA
    • EFA config: efa-config-1.0 (new)
      • Configures hugepages, system limits, and paths for Libfabric and Open MPI installed by EFA

BUG FIXES

2.6.1

ENHANCEMENTS

  • Change ProctrackType from proctrack/gpid to proctrack/cgroup in slurm.conf in order to better handle termination of stray processes when running MPI applications. This also includes the creation of a cgroup Slurm configuration in in order to enable the cgroup plugin.
  • Skip execution, at node bootstrap time, of all those install recipes that are already applied at AMI creation time. The old behaviour can be restored setting the property "skip_install_recipes" to "no" through extra_json. The old behaviour is required in case a custom_node_package is specified and could be needed in case custom_cookbook is used (depending or not if the custom cookbook contains changes into any *_install recipes)
  • Start CloudWatch agent earlier in the node bootstrapping phase so that cookbook execution failures are correctly uploaded and are available for troubleshooting.

CHANGES

  • FSx Lustre: remove x-systemd.requires=lnet.service from mount options in order to rely on default lnet setup provided by Lustre.
  • Enforce Packer version to be >= 1.4.0 when building an AMI. This is also required for customers using pcluster createami command.
  • Remove /tmp/proxy.sh file. Proxy configuration is now written into /etc/profile.d/proxy.sh
  • Omit cfn-init-cmd and cfn-wire from the files stored in CloudWatch logs.

BUG FIXES

  • Fix installation of Intel Parallel Studio XE Runtime that requires yum4 since version 2019.5.
  • Fix compilation of Torque scheduler on Ubuntu 18.04.

2.6.0

ENHANCEMENTS

  • Add support for Amazon Linux 2
  • Install NICE DCV on Ubuntu 18.04 (this includes ubuntu-desktop, lightdm, mesa-util packages)
  • Install and setup Amazon Time Sync on all OSs
  • Compile Slurm with mysql accounting plugin on Ubuntu 18.04 and Ubuntu 16.04
  • Enable FSx Lustre on Ubuntu 18.04 and Ubuntu 16.04

CHANGES

  • Upgrade EFA installer to version 1.8.3:
    • Kernel module: efa-1.5.1 (updated from efa-1.4.1)
    • RDMA core: rdma-core-25.0 (distributed only) (no change)
    • Libfabric: libfabric-aws-1.9.0amzn1.1 (updated from libfabric-aws-1.8.1amzn1.3)
    • Open MPI: openmpi40-aws-4.0.2 (no change)
  • Add SHA256 checksum verification to verify integrity of NICE DCV packages
  • Upgrade Slurm to version 19.05.5
  • Install Python 2.7.17 on CentOS 6 and set it as default through pyenv
  • Install Ganglia from repository on Amazon Linux, Amazon Linux 2, CentOS 6 and CentOS 7
  • Disable StrictHostKeyChecking for SSH client when target host is inside cluster VPC for all OSs except CentOS 6
  • Pin Intel Python 2 and Intel Python 3 to version 2019.4
  • Automatically disable ptrace protection on Ubuntu 18.04 and Ubuntu 16.04 compute nodes when EFA is enabled
  • Packer version >= 1.4.0 is required for AMI creation

BUG FIXES

  • Fix issue with slurmd daemon not being restarted correctly when a compute node is rebooted
  • Fix errors causing Torque not able to locate jobs, setting server_name to fqdn on master node
  • Fix Torque issue that was limiting the max number of running jobs to the max size of the cluster
  • Slurm: configured StateSaveLocation and SlurmdSpoolDir directories to be writable only to slurm user

2.5.1

CHANGES

  • Upgrade NVIDIA driver to Tesla version 440.33.01.
  • Upgrade CUDA library to version 10.2.
  • Upgrade EFA installer to version 1.7.1:
    • Kernel module: efa-1.4.1
    • RDMA core: rdma-core-25.0
    • Libfabric: libfabric-aws-1.8.1amzn1.3
    • Open MPI: openmpi40-aws-4.0.2

BUG FIXES

  • Fix installation of NVIDIA drivers on Ubuntu 18.
  • Fix installation of CUDA toolkit on Centos 6.
  • Fix installation of Munge on Amazon Linux, Centos 6, Centos 7 and Ubuntu 16.
  • Export shared directories to all CIDR blocks in a VPC rather than just the first one.

2.5.0

ENHANCEMENTS

  • Install NICE DCV on Centos 7 (this includes Gnome and Xorg packages).
  • Install Intel Parallel Studio 2019.5 Runtime in Centos 7 AMI and share /opt/intel over NFS.
  • Add support for Ubuntu 18.

CHANGES

  • Remove support for Ubuntu 14.
  • Upgrade Intel MPI to version U5.
  • Upgrade EFA Installer to version 1.6.2, this also upgrades Open MPI to 4.0.2.
  • Upgrade NVIDIA driver to Tesla version 418.87.
  • Upgrade CUDA library to version 10.1.
  • Upgrade Slurm to version 19.05.3-2.
  • Slurm: changed following parameters in global configuration:
    • SelectType=cons_tres, SelectTypeParameter=CR_CPU_Memory, GresTypes=gpu: needed to enable support for GPU scheduling.
    • EnforcePartLimits=ALL: jobs which exceed a partition's size and/or time limits will be rejected at submission time.
    • Removed FastSchedule since deprecated.
    • SlurmdTimeout=180, UnkillableStepTimeout=180: to allow scheduler to recover especially when under heavy load.
  • Echo compute instance type and memory information in COMPUTE_READY message
  • Changes to sshd config:
  • Increase default root volume to 25GB.
  • Enable flock user_xattr noatime Lustre options by default everywhere and x-systemd.automount x-systemd.requires=lnet.service for systemd based systems.
  • Install EFA in China AMIs.

BUG FIXES

  • Fix Ganglia not starting on Ubuntu 16
  • Fix bug that was preventing nodes to mount partitioned EBS volumes.

2.4.1

ENHANCEMENTS

  • Install IntelMPI on Alinux, Centos 7 and Ubuntu 1604
  • Upgrade EFA to version 1.4.1
  • Run all node daemons and cookbook recipes in isolated Python virtualenvs. This allows our code to always run with the required Python dependencies and solves all conflicts and runtime failures that were being caused by user packages installed in the system Python

CHANGES

  • Torque: upgrade to version 6.1.2
  • Run all node daemons with Python 3.6
  • Torque: changed following parameters in global configuration:
    • server node_check_rate = 120 - Specifies the minimum duration (in seconds) that a node can fail to send a status update before being marked down by the pbs_server daemon. Previously was 600. This reduces scaling reaction times in case of instance failure or unexpected termination (especially with spot)
    • server node_ping_rate = 60 - Specifies the maximum interval (in seconds) between successive "pings" sent from the pbs_server daemon to the pbs_mom daemon to determine node/daemon health. Previously was 300. Setting it to half the node_check_rate.
    • server timeout_for_job_delete = 30 - The specific timeout used when deleting jobs because the node they are executing on is being deleted. Previously was 120. This prevents job deletion to hang for more than 30 seconds when the node they are running on is being deleted.
    • server timeout_for_job_requeue = 30 - The specific timeout used when requeuing jobs because the node they are executing on is being deleted. Previously was 120. This prevents node deletion to hang for more than 30 seconds when a job cannot be rescheduled.

BUG FIXES

  • Restore correct value for filehandle_limit that was getting reset when setting memory_limit for EFA
  • Torque: fix configuration of server operators that was preventing compute nodes from disabling themselves before termination

2.4.0

ENHANCEMENTS

  • Add support for EFA on Centos 7, Amazon Linux and Ubuntu 1604
  • Add support for Ubuntu in China region cn-northwest-1

CHANGES

  • SGE: changed following parameters in global configuration
    • max_unheard 00:03:00: allows a faster reaction in case of faulty nodes
    • reschedule_unknown 00:00:30: enables rescheduling of jobs running on failing nodes
    • qmaster_params ENABLE_FORCED_QDEL_IF_UNKNOWN: forces job deletion on unresponsive nodes
    • qmaster_params ENABLE_RESCHEDULE_KILL: forces rescheduling or killing of jobs running on failing nodes
  • Slurm: decrease SlurmdTimeout to 120 seconds to speed up replacement of faulty nodes
  • Always use full master FQDN when mounting NFS on compute nodes. This solves some issues occurring with some networking setups and custom DNS configurations
  • Set soft and hard ulimit on open files to 10000 for all supported OSs
  • Pin python supervisor version to 3.4.0
  • Remove unused compute_instance_type from jobwatcher.cfg
  • Removed unused max_queue_size from sqswatcher.cfg
  • Remove double quoting of the post_install args

BUG FIXES

  • Fix issue that was preventing Torque from being used on Centos 7
  • Start node daemons at the end of instance initialization. The time spent for post-install script and node initialization is not counted as part of node idletime anymore.
  • Fix issue which was causing an additional and invalid EBS mount point to be added in case of multiple EBS
  • Install Slurm libpmpi/libpmpi2 that is distributed in a separate package since Slurm 17

2.3.1

ENHANCEMENTS

  • FSx Lustre - add support in Amazon Linux

CHANGES

  • Slurm - upgrade to version 18.08.6.2
  • Slurm - declare nodes in separate config file and use FUTURE for dummy nodes
  • Slurm - set ReturnToService=1 in scheduler config in order to recover instances that were initially marked as down due to a transient issue.
  • NVIDIA - update drivers to version 418.56
  • CUDA - update toolkit to version 10.0
  • Increase default EBS volume size from 15GB to 17GB
  • Add LocalHostname to COMPUTE_READY events
  • Pin future, retrying and six packages in Ubuntu 14.04
  • Add stackname and max_queue_size to sqswatcher configuration

2.2.1

ENHANCEMENTS

  • FSx Lustre: add support for FSx Lustre in Centos7. In case of custom AMI, FSx Lustre is only supported with Centos 7.5 and Centos 7.6.

CHANGES

  • SGE: allow users to force job deletion
  • Centos7: use official AMI as the base image when building ParallelCluster AMI

BUG FIXES

  • Torque: wait for scheduler initialization before completing compute node setup
  • EBS: fix block device conversion to correctly attach ebs nvme volumes
  • Packer: retrieve aws-cfn-bootstrap-latest package from cn-north-1 in all China regions.
  • RAID: automatically prepend / to raid_shared_dir when not specified in cluster config
  • EBS: remove extra leading / when mounting EBS volumes

TOOLING

  • Add a script to simplify cookbook package upload when using custom_chef_cookbook option

2.1.1

  • China Regions, cn-north-1 and cn-northwest-1 support

2.1.0

  • EFS support
  • RAID support

2.0.2

  • Fix issue with jq on ubuntu1404 and centos6. Now using version 1.4
  • Fix dependency issue with AWS CLI package on ubuntu1404

2.0.0

  • Rename CfnCluster to AWS ParallelCluster
  • Support multiple EBS Volumes
  • Add AWS Batch as a supported scheduler
  • Support Custom AMI's

1.6.0

  • Add scaledown_idletime to nodewatcher config
  • Add cookbook recipes for jobwatcher
  • Remove publish_pending scripts

1.5.4

  • Set SGE Accounting summary to be true, this reports a single accounting record for a mpi job
  • Add option to disable ganglia extra_json = { "cfncluster" : { "ganglia_enabled" : "no" } }

1.5.2

  • Fix bug that prevented c5d/m5d instances from working
  • Set CPU as a consumable resource in slurm config

1.5.1

Major new features/updates:

  • Added parameter to specify custom cfncluster-node package

Bug fixes/minor improvements:

  • Fixed poise-python dependecy issue
  • Poll on EBS Volume attachment status
  • Added more info on failure of pre and post install files
  • Fixed SLURM cron job to publish pending metric

1.4.1

Major new features/updates:

  • Updated to latest cfncluster-node 1.4.3

1.4.0

Major new features/updates:

  • Updated to Amazon Linux 2017.09.1
  • Applied patches to Ubuntu 16.04
  • Applied patches to Ubuntu 14.04
  • Updated to Centos 7.4
  • Upgraded Centos 6 AMI
  • Updated to Nvidia driver 384
  • Updated to CUDA 9
  • Updated to latest cfncluster-node 1.4.2

Bug fixes/minor improvements:

  • Added support for NVMe-based instance store
  • Fixed ganglia plotting issue on ubuntu
  • Fixed slow SLURM scaling times on systemd platforms.

1.3.2

  • Relicensed to Apache License 2.0
  • Updated to Amazon Linux 2017.03
  • Pulled in latest cookbook dependencies
  • Removed Openlava support

1.2.0

  • Dougal Ballantyne
    • Updated to Chef 12.8.1
    • Updated Openlava to 3.1.3
    • Updated SGE to 8.1.9
    • Updated cfncluster-node to 1.1.0
    • Added slots to compute-ready script
    • Updated cookbook dependencies

1.1.0

  • Dougal Ballantyne - Updated to Amazon Linux 2015.09.2 for base AMI

1.0.1

  • Dougal Ballantyne
    • Fix Ganglia rebuild on 2nd run
    • Update to cfncluster-node==1.0.1

1.0.0

  • Dougal Ballantyne - 1.0.0 release of cookbook matching 1.0.0 release of cfncluster.

0.1.0

  • Dougal Ballantyne - Initial release of cfncluster-cookbooks