HFS 2.5.7 Release Notes for HPE Compute Scale-Up
Server 3200 and Superdome Flex Family systems
Linux Operating Systems:
Minimum Linux
Distro versions on Compute Scale-up
Server 3200:
HFS 2.5.7 on Compute
Scale-up Server 3200 supports Linux Distro versions:
HFS 2.5.6 on Compute
Scale-up Server 3200 supports Linux Distro versions:
HFS 2.5.5 on Compute
Scale-up Server 3200 supports Linux Distro versions:
HFS 2.5.4 on Superdome
Flex and Superdome Flex 280 supports
Linux versions.
For
older OS versions, use the following HFS bundles:
Note:
Customers running older distro versions not included in HFS 2.5.7 are still
supported.
Installation:
The following linux bootline options are recommended when installing. HFS
will add the bootline options automatically when
needed:
|
tsc=nowatchdog |
Prevent the watchdog from changing the clocksource
from tsc. |
|
add_efi_memmap |
Ensure all memory is included in Linux memory map. |
|
udev.children-max=512 |
Prevent driver load issues when booting. |
|
nmi_watchdog=0 |
Disable SW watchdog, which may have scaling issues on
large systems. |
|
watchdog_thresh=30 |
Increase timeouts on large systems. 32 socket systems may
require 60 seconds. |
|
workqueue.watchdog_thresh=60 |
Increase timeouts on large systems. 32 sockets systems may
require 120 seconds. |
|
pci=nobar |
Prevent Linux from assigning unassigned BARs. |
|
pci=norom |
Do not assign address space to expansion ROMs that do not
already have BIOS assigned address ranges. |
|
console=ttyS0,115200 |
Enable the serial console. |
|
earlyprintk=ttyS0,115200 |
Display early boot messages. Aids in debugging early
boot issues. |
Note:
Removing "quiet" from the kernel bootline
will also aid debugging boot issues.
Linux
distro links:
HFS (HPE Foundation Software) Description:
HPE Foundation Software (HFS) includes automatic boot-time optimization utilities,
reliability features, and technical support tools. Designed for high
performance computing, these tools help maximize system performance and
availability.
HPE Documentation Links:
HPE
Foundation Software (HFS 2.5.7) on HPE Support Center
HPE
Foundation Software (HFS 2.5.7) on Software Download Repository
HPE
Compute Scale-up Server 3200 Linux Installation Guide
HPE
Compute Scale-up Server 3200 Quick Specs
HPE
Superdome Flex Administrator Guide
Managing
System Performance with HPE Foundation Software
HPE
Foundation Software (HFS) commands
PRODUCT MODEL(S):
HPE Compute Scale-Up Server 3200
HPE Superdome Flex
HPE Superdome Flex 280
HFS 2.5.7
ENHANCEMENTS / FIXES:
·
New distro support:
o
SLES 16
o
RHEL 10.1, 9.7
·
libcpuset:
o
sles16 uses cpuset V2. Add libcpuset_compat to provide a compatibility layer for
applications using cpusets.
·
numatools:
o
sles16 uses cpuset V2.
Use libcpuset_compat to provide a compatibility
layer for applications using cpusets.
·
System:
·
topology:
·
dcd:
o
Update DCD to version 4.9-5.1
SUPERSEDES:
Version: HFS 2.5.6
UPDATE RECOMMENDATION:
Recommended
Notes:
·
SLES16 uses cgroup V2 and does not include V1.
·
The Pilot4 Linux /sys/device
resource0 file, when opened opened (as root)
and mapped (mmap()), consecutive reads or writes can generate PIC errors
which can crash the system. By default, the resource0
file has root only rw permission, so non-root cannot
hit this issue. Also, the file needs to be mapped, meaning commands such
as cat will not hit the issue. The change is to remove rw permission on the resource0 file, to
prevent root programs from mapping the file and hitting the issue. root
can change permission back to rw if for some reason they
want to, though that exposes the issue which can crash the system. Pilot4
is on Superdome Flex and Fles280 platforms. It is not on SUS 3200
platform.
For
more details and workaround, see Customer Advisory a00108899
INSTALLATION
INSTRUCTIONS:
Please review all instructions and the
"Hewlett Packard Enterprise Support Tool License Terms" or your
Hewlett Packard Enterprise support terms and conditions for precautions, scope
of license, restrictions, and limitation of liability and warranties, before
installing this package. It is important that you read and understand these
instructions completely before you begin. This can determine your success in
completing the software update.
Linux Installation instructions:
Installation Notes:
·
Note:
Sles15 sp4 kernel version 5.14.21-150400.26.63
has a regression. It does not advertise avx capabilities, causing poor performance by applications
wanting to use avx instructions. Fixed in version 5.14.21-150400.26.66
. SUSE BZ 1211205 for
details.
HFS ISO installation Instructions:
Note: To see the list of
groups to install, use: dnf grouplist
f. Reboot the system to activate the change:
g.
reboot
After the system reaches the EFI shell, in the RMC command window, enter:
power reset npar pnum=0
Note: Refer to your operating system
documentation for details on adding directories of RPM packages as available
software sources/repositories for use by zipper, dnf
and yum.
Installing HFS in
a contaner:
These
directions are in the hfs-container rpm README file.
#
These are example commands to build a container, install
HPE-HFS, and run the resulting container on a
# HPE Scale Up Server 3200 worker node in a Red Hat Openshift
Platform 4.14 cluster. This solution requires
# the Dockerfile and hfs-config.repo file included in the hfs-container
rpm.
#
# This solution also requires the HPE HFS iso file,
hpe-foundation-2.5.7-cd1-media-rhel97-x86_64.iso. One
# method used to provide the iso contents to the container build is a
loop mount command shown below.
#
mkdir ./HPE-Foundation-Software-2.5.7
sudo mount -o loop ./hpe-foundation-2.5.7-cd1-media-rhel97-x86_64.iso
./HPE-Foundation-Software-2.5.7
#
# Commands to extract the hfs-container
files from the HPE-HFS iso image
#
rpm2cpio HPE-Foundation-Software-2.5.7/RPMS/hfs-container*.rpm
| cpio -icd
cp opt/hpe/container/* .
#
# With copies of the Dockerfile and hfs-config.repo, and the iso file mounted as show above,
the command
# below will build a local image using the definitions in Dockerfile.
#
sudo podman image
build -f Dockerfile -t hfs:1
#
# The resulting container can be tested for basic functionality on
the build host with the folllowing
# command.
#
sudo podman run
-it --network host --privileged --pid host --volume
/:/host:ro localhost/hfs:1
#
# The commands below were developed from the Red Hat OpenShift
documentation provided in the link below
#
#
https://docs.openshift.com/container-platform/4.13/registry/securing-exposing-registry.html#registry-exposing-hpe-hfs-registry-manually_securing-exposing-registry
#
#
# Login to the Openshift cluster
#
oc login https://api.${CLUSTER-HOST}:6443
-u kubeadmin -p ${PASSWORD}
#
# Get the default route to the internal OpenShift registry::
#
HOST=$(oc get route default-route -n openshift-image-registry --template='{{ .spec.host }}')
#
# Get the certificate of the Ingress Operator on the internal Openshift registry:
#
oc get secret -n openshift-ingress
router-certs-default -o go-template='{{index .data "tls.crt"}}' |
base64 -d | sudo tee /etc/pki/ca-trust/source/anchors/${HOST}.crt
> /dev/null
#
# Enable the cluster’s default certificate to trust the route using the
following commands:
#
sudo update-ca-trust enable
#
# Log in with podman using the default route:
#
sudo podman login
-u kubeadmin -p $(oc whoami -t) $HOST
#
# Tag the local image with the destination project/imagestream:version
#
sudo podman tag
localhost/hfs:1 ${HOST}/hpe-hfs/hfs:version1
#
# Create the hpe-hfs namespace for the rest of the
process below
#
oc create namespace hpe-hfs
#
# Change to hpe-hfs namespace
#
oc project hpe-hfs
#
# Add permission to run privileged containers and pods from daemonsets
#
oc adm policy
add-scc-to-user privileged -z default -n hpe-hfs
#
# Check if the destination imagesource already exists
#
oc describe imagestream
hfs -n hpe-hfs
#
# If the imagestream does not exist, create a new one
#
oc create imagestream
hfs -n hpe-hfs
#
# Push the image from the local podman registry to
the internal Openshift registry
#
sudo podman push
${HOST}/hpe-hfs/hfs:version1
#
# Check the destination imagestream
#
oc describe imagestream
hfs -n hpe-hfs
#
# Run a test pod using the new image
#
oc apply -f ./hfs.yaml
#
# Check the status of the pod
#
oc get pods
#
# rsh in to the pod to manually change the
configuration and execute the test
#
oc rsh hfs-test
#
# The following file is an example of tuning suggestions for the HPE Scale Up
Server 3200
# This example has 8 sockets and divides each socket in half for reserved cores
for general
# purpose, OS, and OpenShift control plane workloads. The other half are
isolated for
# low latency and high performance workloads. The hugepages
section is also an example.
#
10-profile-hpe-auto-config.yaml
DISCLAIMER:
The information in this document is subject to change without notice.
Hewlett Packard Enterprise makes no warranty of any
kind with regard to this material, including, but not limited to, the implied
warranties of merchantability and fitness for a particular purpose. Hewlett
Packard Enterprise shall not be liable for errors contained herein or for
incidental or consequential damages in connection with the furnishing,
performance, or use of this material.
This document contains
proprietary information that is protected by copyright. All rights are
reserved. No part of this document may be reproduced, photocopied, or
translated to another language without the prior written consent of Hewlett
Packard Enterprise.
(C) Copyright 2025 Hewlett
Packard Enterprise Development L.P.