Provisioning of bare-metal servers

The manual installation of a system is described below.

The use of an installation server is recommended. Cobbler, which is integrated in OSISM, can be used for this purpose.

When using Cobbler it is only necessary to manually prepare the bare-metal system used as manager node. The remaining nodes can then be provisioned using Cobbler.

The manual node installation is completely possible without network connectivity.


  • Download the latest ISO image for Ubuntu 18.04 from

    • Use the ubuntu-18.04.2-server-amd64.iso image

    • Do not use the ubuntu-18.04.2-live-server-amd64.iso image

    • The version number may be different, always use the latest available version of 18.04 LTS

  • Create a bootable USB stick from this ISO image. Alternatively you can also work with a CD

  • Perform a hardware RAID configuration if necessary

  • Boot bare-metal server from this USB stick/CD


  • Choose English as language

  • Choose Install Ubuntu Server

  • Choose English as language (again)

  • Choose your location (e.g. Germany)

  • Choose en_US.UTF-8 as locale

  • Choose the keyboard layout from a list, use English (US)

  • Choose and configure the primary network interface

    • Depending on the environment, the network may not work at this point. Then select any interface and then select Do not configure the network at this time in the next step.

  • Set the hostname (the hostname is 60-10 and not

  • Set ubuntu as full name for the new user

  • Set ubuntu as the username for the account

    • The later used operator user dragon is created during the bootstrap and should not be created during the installation.

  • Set a password for the account

    • The account is only needed initially and can be deleted after completion of the bootstrap.

  • Choose Manual as partitioning method and execute the partitioning according to company specifications

  • Choose No automatic updates

  • Choose OpenSSH server as software to install

  • After completion, restart the system


python-minimal must be installed on the systems.


  • The use of a UEFI is recommended

  • The use of a RAID is recommended

  • The use of a LVM2 is recommended

  • The use of own file systems for the following mountpoints is recommended

    • / (10 GByte)

    • /home (2 GByte)

    • /tmp (2 GByte)

    • /var/lib/docker (30 GByte, do not set the nosuid flag on /var/lib/docker)

    • /var/log/audit (1 GByte)

    • /var (10 GByte)


    The size of the individual partitions is minimal. Depending on the node type, the individual partitions should be made larger. This applies in particular to /var/lib/docker. On controllers at least 100 GByte should be used.

    A later enlargement is possible during operation.

    # lvextend -L +10G /dev/mapper/system-docker
    # resize2fs -p /dev/mapper/system-docker
  • The use of a swap partition with at least 8 GByte is recommended



When using XFS as the file system for /var/lib/docker, note the following: Running on XFS without d_type support now causes Docker to skip the attempt to use the overlay or overlay2 driver.


EFI partitions

# lsblk
sda                   8:0    0 59.6G  0 disk
├─sda1                8:1    0  476M  0 part  /boot/efi
└─sda2                8:2    0 59.2G  0 part
  └─md0               9:0    0 59.1G  0 raid1
    ├─system-root   253:0    0  9.3G  0 lvm   /
    ├─system-swap   253:1    0  7.5G  0 lvm   [SWAP]
    ├─system-tmp    253:2    0  1.9G  0 lvm   /tmp
    ├─system-audit  253:3    0  952M  0 lvm   /var/log/audit
    ├─system-var    253:4    0  9.3G  0 lvm   /var
    ├─system-docker 253:5    0  9.3G  0 lvm   /var/lib/docker
    └─system-home   253:6    0  1.9G  0 lvm   /home
sdb                   8:16   0 59.6G  0 disk
├─sdb1                8:17   0  476M  0 part
└─sdb2                8:18   0 59.2G  0 part
  └─md0               9:0    0 59.1G  0 raid1
    ├─system-root   253:0    0  9.3G  0 lvm   /
    ├─system-swap   253:1    0  7.5G  0 lvm   [SWAP]
    ├─system-tmp    253:2    0  1.9G  0 lvm   /tmp
    ├─system-audit  253:3    0  952M  0 lvm   /var/log/audit
    ├─system-var    253:4    0  9.3G  0 lvm   /var
    ├─system-docker 253:5    0  9.3G  0 lvm   /var/lib/docker
    └─system-home   253:6    0  1.9G  0 lvm   /home
# dd if=/dev/sda1 of=/dev/sdb1
# efibootmgr -v | grep ubuntu
Boot0000* ubuntu     HD(1,GPT,f6b80cef-a636-439a-b2c2-e30bc385eada,0x800,0xee000)/File(\EFI\UBUNTU\SHIMX64.EFI)
Boot0018* ubuntu     HD(1,GPT,f6b80cef-a636-439a-b2c2-e30bc385eada,0x800,0xee000)/File(\EFI\UBUNTU\GRUBX64.EFI)
# efibootmgr -c -d /dev/sdb -p 1 -L "ubuntu2" -l "\EFI\UBUNTU\GRUBX64.EFI"
# efibootmgr -c -d /dev/sdb -p 1 -L "ubuntu2" -l "\EFI\UBUNTU\SHIMX64.EFI"


After the first boot depending on the environment it is necessary to create the network configuration for the management interface manually, because for example bonding or VLANs should be used.

  • At the beginning it is sufficient to be able to reach the system via SSH.

  • It is not necessary to create the entire network configuration. The network configuration is created during the bootstrap on the systems.


# modprobe bonding
# ip link add bond0 type bond
# ip link set bond0 type bond miimon 100 mode 802.3ad lacp_rate 1
# ip link set eno1 down
# ip link set eno1 master bond0
# ip link set eno2 down
# ip link set eno2 master bond0
# ip link set bond0 up
# cat /proc/net/bonding/bond0
# ip link add link bond0 name vlan101 type vlan id 101
# ip link set vlan101 up
# ip address add dev vlan101
# ip route add default via
  • You may have to set the nameservers in /etc/resolv.conf. Temporarily remove the entry.


  version: 2
  renderer: networkd
      dhcp4: no
      dhcp4: no
      dhcp4: no
        - eno1
        - eno2
        mode: 802.3ad
        lacp-rate: fast
        mii-monitor-interval: 100
      id: 101
      link: bond0
      addresses: [ "" ]
       - to:
        search: [ ]
        addresses: [ "", "" ]
# netplan apply