Pre-Installation

The page walks a user through setting up the Cray LiveCD with the intention of installing Cray System Management (CSM).

  1. Boot installation environment
    1. Prepare installation environment server
    2. Boot the LiveCD
    3. First log in
    4. Prepare the data partition
    5. Set reusable environment variables
    6. Exit the console and log in with SSH
  2. Import CSM tarball
    1. Download CSM tarball
    2. Import tarball assets
  3. Create system configuration
    1. Validate SHCD
    2. Generate topology files
    3. Customize system_config.yaml
    4. Run CSI
    5. Prepare Site Init
    6. Initialize the LiveCD
  4. Next topic

1. Boot installation environment

This section walks the user through booting and connecting to the LiveCD.

Before proceeding, the user must obtain the CSM tarball containing the LiveCD.

NOTE: Each step denotes where its commands must run; external# refers to a server that is not the Cray, whereas pit# refers to the LiveCD itself.

Any steps run on an external server require that server to have the following tools:

  • ipmitool
  • ssh
  • tar

NOTE: The CSM tarball will be fetched from the external server in the Import tarball assets step using curl or scp. If a web server is not installed, then scp is the backup option.

1.1 Prepare installation environment server

  1. (external#) Download the CSM software release from the public Artifactory instance.

    NOTES:

    • -C - is used to allow partial downloads. These tarballs are large; in the event of a connection disruption, the same curl command can be used to continue the disrupted download.
    • If air-gapped or behind a strict firewall, then the tarball must be obtained from the medium delivered by Cray-HPE. For these cases, copy or download the tarball to the working directory and then proceed to the next step. The tarball will need to be fetched with scp during the Download CSM tarball step.

  2. (external#) Set the CSM RELEASE version

    Example release versions:

    • An alpha build: CSM_RELEASE=1.3.0-alpha.99
    • A release candidate: CSM_RELEASE=1.3.0-rc.1
    • A stable release: CSM_RELEASE=1.3.0
    CSM_RELEASE=<value>

  3. (external#) Download the CSM tarball

    NOTE: CSM does NOT support the use of proxy servers for anything other than downloading artifacts from external endpoints. Using http_proxy or https_proxy in any way other than the following examples will cause many failures in subsequent steps.

    • Without proxy:

      curl -C - -f -O "https://release.algol60.net/$(awk -F. '{print "csm-"$1"."$2}' <<< ${CSM_RELEASE})/csm/csm-${CSM_RELEASE}.tar.gz"

    • With https proxy:

      https_proxy=https://example.proxy.net:443 curl -C - -f -O "https://release.algol60.net/$(awk -F. '{print "csm-"$1"."$2}' <<< ${CSM_RELEASE})/csm/csm-${CSM_RELEASE}.tar.gz"

  4. (external#) Extract the LiveCD from the tarball.

    OUT_DIR="$(pwd)/csm-temp"
mkdir -pv "${OUT_DIR}"
tar -C "${OUT_DIR}" --wildcards --no-anchored --transform='s/.*\///' -xzvf "csm-${CSM_RELEASE}.tar.gz" 'cray-pre-install-toolkit-*.iso'

1.2 Boot the LiveCD

  1. (external#) Start a typescript and set the PS1 variable to record timestamps.

    NOTE: Typescripts help triage if problems are encountered.

    script -a "boot.livecd.$(date +%Y-%m-%d).txt"
export PS1='\u@\H \D{%Y-%m-%d} \t \w # '

  2. (external#) Follow one of the procedures below based on the vendor for the ncn-m001 node:

    • HPE iLO BMCs

      Prepare a server on the network to host the cray-pre-install-toolkit ISO file, if the current server is insufficient. Then follow the HPE iLO BMCs to boot the RemoteISO before returning here.

    • Gigabyte BMCs and Intel BMCs

      Create a USB stick using the following procedure.

      1. (external#) Get cray-site-init from the tarball.

        OUT_DIR="$(pwd)/csm-temp"
mkdir -pv "${OUT_DIR}"
tar -C "${OUT_DIR}" --wildcards --no-anchored --transform='s/.*\///' -xzvf "csm-${CSM_RELEASE}.tar.gz" 'cray-site-init-*.rpm'

      2. (external#) Install the write-livecd.sh script:

        • RPM-based systems:

          rpm -Uvh --force ${OUT_DIR}/cray-site-init*.rpm

        • Non-RPM-based systems (requires bsdtar):

          bsdtar xvf "${OUT_DIR}"/cray-site-init-*.rpm --include *write-livecd.sh -C "${OUT_DIR}"
mv -v "${OUT_DIR}"/usr/local/bin/write-livecd.sh "./${OUT_DIR}"
rmdir -pv "${OUT_DIR}/usr/local/bin/"

        • Non-RPM-based distros (requires rpm2cpio):

          rpm2cpio cray-site-init-*.rpm | cpio -idmv
mv -v ./usr/local/bin/write-livecd.sh "./${OUT_DIR}"
rm -vrf ./usr

      3. Follow Bootstrap a LiveCD USB and then return here.

1.3 First log in

On first login, the LiveCD will prompt the administrator to change the password.

  1. (pit#) Log in.

    NOTE: The initial password is empty.

    At the login prompt, enter root as the username. Because the initial password is blank, press return twice at the first two password prompts. The LiveCD will force a new password to be set.

    Password:           <-------just press Enter here for a blank password
You are required to change your password immediately (administrator enforced)
Changing password for root.
Current password:   <------- press Enter here, again, for a blank password
New password:       <------- type new password
Retype new password:<------- retype new password
Welcome to the CRAY Pre-Install Toolkit (LiveOS)

  2. (pit#) Configure the site-link (lan0), DNS, and gateway IP addresses.

    NOTE: The site_ip, site_gw, and site_dns values must come from the local network administration or authority.

    1. Set site_ip variable.

      Set the site_ip value in CIDR format (A.B.C.D/N):

      site_ip=<IP CIDR>

    2. Set the site_gw and site_dns variables.

      Set the site_gw and site_dns values in IPv4 dotted decimal format (A.B.C.D):

      site_gw=<Gateway IP address>
site_dns=<DNS IP address>

    3. Set the site_nics variable.

      The site_nics value or values are found while the user is in the LiveCD (for example, site_nics='p2p1 p2p2 p2p3' or site_nics=em1).

      site_nics='<site NIC or NICs>'

    4. Set the SYSTEM_NAME variable.

      SYSTEM_NAME is the name of the system. This will only be used for the PIT hostname. This variable is capitalized because it will be used in a subsequent section.

      SYSTEM_NAME=<system name>

    5. Run the csi-setup-lan0.sh script to set up the site link and set the hostname.

      NOTES:

      • All of the /root/bin/csi-* scripts can be run without parameters to display usage statements.
      • The hostname is auto-resolved based on reverse DNS.
      /root/bin/csi-setup-lan0.sh "${site_ip}" "${site_gw}" "${site_dns}" "${site_nics}"

  3. (pit#) Verify that the assigned IP address was successfully applied to lan0 .

    wicked ifstatus --verbose lan0

    NOTE: The output from the above command must say leases: ipv4 static granted. If the IPv4 address was not granted, then go back and recheck the variable values. The output will indicate the IP address failed to assign, which can happen if the given IP address is already taken on the connected network.

1.4 Prepare the data partition

  1. (pit#) Mount the PITDATA partition.

    Use either the RemoteISO or the USB option below, depending how the LiveCD was connected in the Boot the LiveCD step.

    • RemoteISO

      Use a local disk for PITDATA:

      disk="$(lsblk -l -o SIZE,NAME,TYPE,TRAN -e7 -e11 -d -n | grep -v usb | sort -h | awk '{print $2}' | xargs -I {} bash -c "if ! grep -Fq {} /proc/mdstat; then echo {}; fi" | head -n 1)"
echo "Using ${disk}"
parted --wipesignatures -m --align=opt --ignore-busy -s "/dev/${disk}" -- mklabel gpt mkpart primary ext4 2048s 100%
partprobe "/dev/${disk}"
mkfs.ext4 -L PITDATA "/dev/${disk}1"
mount -vL PITDATA

    • USB

      Mount the USB data partition:

      mount -vL PITDATA

1.5 Set reusable environment variables

These variables will need to be set for many procedures within the CSM installation process.

NOTE: This sets some variables that were already set. These should be set again anyway.

  1. (pit#) Set the variables.

    1. Set the PITDATA variable.

      export PITDATA="$(lsblk -o MOUNTPOINT -nr /dev/disk/by-label/PITDATA)"

    2. Set the CSM_RELEASE variable.

      The value is based on the version of the CSM release being installed.

      Example release versions:

      • An alpha build: CSM_RELEASE=1.3.0-alpha.99
      • A release candidate: CSM_RELEASE=1.3.0-rc.1
      • A stable release: CSM_RELEASE=1.3.0
      export CSM_RELEASE=<value>

    3. Set the CSM_PATH variable.

      After the CSM release tarball has been expanded, this will be the path to its base directory.

      export CSM_PATH="${PITDATA}/csm-${CSM_RELEASE}"

    4. Set the SYSTEM_NAME variable.

      This is the user friendly name for the system. For example, for eniac-ncn-m001, SYSTEM_NAME should be set to eniac.

      export SYSTEM_NAME=<value>

  2. (pit#) Update /etc/environment.

    This ensures that these variables will be set in all future shells on the PIT node.

    cat << EOF >/etc/environment
CSM_RELEASE=${CSM_RELEASE}
CSM_PATH=${PITDATA}/csm-${CSM_RELEASE}
GOSS_BASE=${GOSS_BASE}
PITDATA=${PITDATA}
SYSTEM_NAME=${SYSTEM_NAME}
EOF

1.6 Exit the console and log in with SSH

  1. (pit#) Create the admin directory for the typescripts and administrative scratch work.

    mkdir -pv "$(lsblk -o MOUNTPOINT -nr /dev/disk/by-label/PITDATA)/prep/admin"
ls -l "$(lsblk -o MOUNTPOINT -nr /dev/disk/by-label/PITDATA)/prep/admin"

  2. (pit#) Exit the typescript and log out.

    exit
exit

  3. (pit#) Exit the console.

    This is done by typing the key-sequence: tilde, period. That is, ~.

    If the console was accessed over an SSH session (that is, the user used SSH to log into another server, and from there used ipmitool to access the console), then press tilde twice followed by a period, in order to prevent exiting the parent SSH session. That is, ~~.

  4. (external#) Copy the typescript to the running LiveCD.

    scp boot.livecd.*.txt root@eniac-ncn-m001:/tmp/

  5. (pit#) SSH into the LiveCD.

    livecd=eniac-ncn-m001.example.company.com
ssh root@"${livecd}"

  6. (pit#) Copy the previous typescript and start a new one.

    cp -pv /tmp/boot.livecd.*.txt "${PITDATA}/prep/admin"
script -af "${PITDATA}/prep/admin/csm-install.$(date +%Y-%m-%d).txt"
export PS1='\u@\H \D{%Y-%m-%d} \t \w # '

  7. (pit#) Print information about the booted PIT image for logging purposes.

    Having this information in the typescript can be helpful if problems are encountered during the install.

    /root/bin/metalid.sh

    Expected output looks similar to the following (the versions in the example below may differ). There should be no errors.

    = PIT Identification = COPY/CUT START =======================================
VERSION=1.6.0
TIMESTAMP=20220504161044
HASH=g10e2532
2022/05/04 17:08:19 Using config file: /var/www/ephemeral/prep/system_config.yaml
CRAY-Site-Init build signature...
Build Commit   : 0915d59f8292cfebe6b95dcba81b412a08e52ddf-main
Build Time     : 2022-05-02T20:21:46Z
Go Version     : go1.16.10
Git Version    : v1.9.13-29-g0915d59f
Platform       : linux/amd64
App. Version   : 1.17.1
metal-ipxe-2.2.6-1.noarch
metal-net-scripts-0.0.2-20210722171131_880ba18.noarch
metal-basecamp-1.1.12-1.x86_64
pit-init-1.2.20-1.noarch
pit-nexus-1.1.4-1.x86_64
= PIT Identification = COPY/CUT END =========================================

2. Import CSM tarball

2.1 Download CSM tarball

  1. Download and install the latest documentation and scripts RPMs, see Check for latest documentation.

  2. (pit#) Download the CSM tarball.

    • From Cray using curl:

      • -C - is used to allow partial downloads. These tarballs are large; in the event of a connection disruption, the same curl command can be used to continue the disrupted download.
      • CSM does NOT support the use of proxy servers for anything other than downloading artifacts from external endpoints. Using http_proxy or https_proxy in any way other than the following examples will cause many failures in subsequent steps.

      • Without proxy:

        curl -C - -f -o "/var/www/ephemeral/csm-${CSM_RELEASE}.tar.gz" \
"https://release.algol60.net/$(awk -F. '{print "csm-"$1"."$2}' <<< ${CSM_RELEASE})/csm/csm-${CSM_RELEASE}.tar.gz"

      • With HTTPS proxy:

        https_proxy=https://example.proxy.net:443 curl -C - -f -o "/var/www/ephemeral/csm-${CSM_RELEASE}.tar.gz" \
"https://release.algol60.net/$(awk -F. '{print "csm-"$1"."$2}' <<< ${CSM_RELEASE})/csm/csm-${CSM_RELEASE}.tar.gz"

    • scp from the external server used in Prepare installation environment server:

      scp "<external-server>:/<path>/csm-${CSM_RELEASE}.tar.gz" /var/www/ephemeral/

2.2 Import tarball assets

If resuming at this stage, the CSM_RELEASE and PITDATA variables are already set in /etc/environment from the Download CSM tarball step.

  1. (pit#) Extract the tarball.

    tar -zxvf  "${PITDATA}/csm-${CSM_RELEASE}.tar.gz" -C ${PITDATA}

  2. (pit#) Install/update the RPMs necessary for the CSM installation.

    NOTE --no-gpg-checks is used because the repository contained within the tarball does not provide a GPG key.

    1. Update cray-site-init.

      NOTE This provides csi, a tool for creating and managing configurations, as well as orchestrating the handoff and deploy of the final non-compute node.

      zypper --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp2/" --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp3/" \
       --no-gpg-checks update -y cray-site-init

    2. Install hpe-csm-goss-package RPM.

      NOTE This package provides the necessary test driver for validating the pre-installation, installation, and more.

      zypper --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp2/" --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp3/" \
       --no-gpg-checks install -y hpe-csm-goss-package

    3. Install csm-testing RPM.

      NOTE This package provides the necessary tests and their dependencies for validating the pre-installation, installation, and more.

      zypper --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp2/" --plus-repo "${CSM_PATH}/rpm/cray/csm/sle-15sp3/" \
       --no-gpg-checks install -y csm-testing

  3. (pit#) Get the artifact versions.

    KUBERNETES_VERSION="$(find ${CSM_PATH}/images/kubernetes -name '*.squashfs' -exec basename {} .squashfs \; | awk -F '-' '{print $NF}')"
echo "${KUBERNETES_VERSION}"
CEPH_VERSION="$(find ${CSM_PATH}/images/storage-ceph -name '*.squashfs' -exec basename {} .squashfs \; | awk -F '-' '{print $NF}')"
echo "${CEPH_VERSION}"

  4. (pit#) Copy the NCN images from the expanded tarball.

    NOTE This hard-links the files to do this copy as fast as possible, as well as to mitigate space waste on the USB stick.

    mkdir -pv "${PITDATA}/data/k8s/" "${PITDATA}/data/ceph/"
rsync -rltDP --delete "${CSM_PATH}/images/kubernetes/" --link-dest="${CSM_PATH}/images/kubernetes/" "${PITDATA}/data/k8s/${KUBERNETES_VERSION}"
rsync -rltDP --delete "${CSM_PATH}/images/storage-ceph/" --link-dest="${CSM_PATH}/images/storage-ceph/" "${PITDATA}/data/ceph/${CEPH_VERSION}"

  5. (pit#) Modify the NCN images with SSH keys and root passwords.

    The following substeps provide the most commonly used defaults for this process. For more advanced options, see Set NCN Image Root Password, SSH Keys, and Timezone on PIT Node.

    1. Generate SSH keys.

      NOTE The code block below assumes there is an RSA key without a passphrase. This step can be customized to use a passphrase if desired.

      ssh-keygen -N "" -t rsa

    2. Export the password hash for root that is needed for the ncn-image-modification.sh script.

      This will set the NCN root user password to be the same as the root user password on the PIT.

      export SQUASHFS_ROOT_PW_HASH="$(awk -F':' /^root:/'{print $2}' < /etc/shadow)"

    3. Inject these into the NCN images by running ncn-image-modification.sh from the CSM documentation RPM.

      NCN_MOD_SCRIPT=$(rpm -ql docs-csm | grep ncn-image-modification.sh)
echo "${NCN_MOD_SCRIPT}"
"${NCN_MOD_SCRIPT}" -p \
   -d /root/.ssh \
   -k "/var/www/ephemeral/data/k8s/${KUBERNETES_VERSION}/kubernetes-${KUBERNETES_VERSION}.squashfs" \
   -s "/var/www/ephemeral/data/ceph/${CEPH_VERSION}/storage-ceph-${CEPH_VERSION}.squashfs"

  6. (pit#) Log the currently installed PIT packages.

    Having this information in the typescript can be helpful if problems are encountered during the install. This command was run once in a previous step – running it again now is intentional.

    /root/bin/metalid.sh

    Expected output looks similar to the following (the versions in the example below may differ). There should be no errors.

    = PIT Identification = COPY/CUT START =======================================
VERSION=1.6.0
TIMESTAMP=20220504161044
HASH=g10e2532
2022/05/04 17:08:19 Using config file: /var/www/ephemeral/prep/system_config.yaml
CRAY-Site-Init build signature...
Build Commit   : 0915d59f8292cfebe6b95dcba81b412a08e52ddf-main
Build Time     : 2022-05-02T20:21:46Z
Go Version     : go1.16.10
Git Version    : v1.9.13-29-g0915d59f
Platform       : linux/amd64
App. Version   : 1.17.1
metal-ipxe-2.2.6-1.noarch
metal-net-scripts-0.0.2-20210722171131_880ba18.noarch
metal-basecamp-1.1.12-1.x86_64
pit-init-1.2.20-1.noarch
pit-nexus-1.1.4-1.x86_64
= PIT Identification = COPY/CUT END =========================================

3. Create system configuration

This stage walks the user through creating the configuration payload for the system.

Run the following steps before starting any of the system configuration procedures.

  1. (pit#) Make the prep directory.

    mkdir -pv "${PITDATA}/prep"

  2. (pit#) Change into the prep directory.

    cd "${PITDATA}/prep"

3.1 Validate SHCD

  1. (pit#) Download the SHCD to the prep directory.

    This will need to be retrieved from the administrator’s Cray deliverable.

  2. Validate the SHCD.

    See Validate SHCD and then return to this page.

3.2 Generate topology files

The following steps use the new, automated method for generating files. The previous step for validate SHCD generated “paddle” files; these are necessary for generating the rest of the seed files.

NOTE The paddle files are temporarily not used due to bugs in the seed file generation software. Until these bugs are resolved, the seed files must be manually generated.

If seed files from a prior installation of the same major-minor version of CSM exist, then these can be used and this step may be skipped.

  1. (pit#) Create each seed file, unless they already exist from a previous installation.

  2. (pit#) Confirm that the following files exist.

    ls -l "${PITDATA}"/prep/{application_node_config.yaml,cabinets.yaml,hmn_connections.json,ncn_metadata.csv,switch_metadata.csv}

    Expected output may look like:

    -rw-r--r-- 1 root root  146 Jun  6 00:12 /var/www/ephemeral/prep/application_node_config.yaml
-rw-r--r-- 1 root root  392 Jun  6 00:12 /var/www/ephemeral/prep/cabinets.yaml
-rwxr-xr-x 1 root root 3768 Jun  6 00:12 /var/www/ephemeral/prep/hmn_connections.json
-rw-r--r-- 1 root root 1216 Jun  6 00:12 /var/www/ephemeral/prep/ncn_metadata.csv
-rw-r--r-- 1 root root  150 Jun  6 00:12 /var/www/ephemeral/prep/switch_metadata.csv

3.3 Customize system_config.yaml

  1. (pit#) Create or copy system_config.yaml.

    • If one does not exist from a prior installation, then create an empty one:

      csi config init empty

    • Otherwise, copy the existing system_config.yaml file into the working directory and proceed to the Run CSI step.

  2. (pit#) Edit the system_config.yaml file with the appropriate values.

    NOTES

    • For a short description of each key in the file, run csi config init --help. IMPORTANT install-ncn-bond-members have many possibilities but are typically:
      • p1p1,p10p1 for HPE nodes.
      • p1p1,p1p2 for Gigabyte nodes.
      • p801p1,p801p2 for Intel nodes.
    • For more details on these settings and the default values, see Default IP Address Ranges and the other topics in CSM Overview.
    • To enable or disable audit logging, refer to Audit Logs for more information.
    • If the system is using a cabinets.yaml file, be sure to update the cabinets-yaml field with 'cabinets.yaml' as its value.
    vim system_config.yaml

3.4 Run CSI

  1. (pit#) Generate the initial configuration for CSI.

    This will validate whether the inputs for CSI are correct.

    csi config init

3.5 Prepare Site Init

Follow the Prepare Site Init procedure.

3.6 Initialize the LiveCD

NOTE: If starting an installation at this point, then be sure to copy the previous prep directory back onto the system.

  1. (pit#) Initialize the PIT.

    The pit-init.sh script will prepare the PIT server for deploying NCNs.

    /root/bin/pit-init.sh

  2. (pit#) Set the IPMI_PASSWORD variable.

    read -r -s -p "NCN BMC root password: " IPMI_PASSWORD

  3. (pit#) Export the IPMI_PASSWORD variable.

    export IPMI_PASSWORD

  4. (pit#) Setup links to the boot artifacts extracted from the CSM tarball.

    NOTES

    • This will also set all the BMCs to DHCP.
    • Changing into the $HOME directory ensures the proper operation of the script.
    cd $HOME && /root/bin/set-sqfs-links.sh

  5. (pit#) Verify that the LiveCD is ready by running the preflight tests.

    csi pit validate --livecd-preflight

    If any tests fail, they need to be investigated. After actions have been taken to rectify the tests (for example, editing configuration or CSI inputs), then restart from the beginning of the Initialize the LiveCD procedure.

  6. Save the prep directory for re-use.

    This needs to be copied off the system and either stored in a secure location or in a secured Git repository. There are secrets in this directory that should not be accidentally exposed.

Next topic

After completing this procedure, proceed to configure the management network switches.

See Configure management network switches.