Bare Metal Server

Available in VPC

This guide describes how to create and manage a Bare Metal server from the NAVER Cloud Platform console.

View server information

You can view the Bare Metal server information in the same way as viewing regular server information. For more information, see Check server information.

Create general server

To use a Bare Metal server, you must create one from the NAVER Cloud Platform console. To create a Bare Metal server, follow these steps:

1. In the VPC environment of the NAVER Cloud Platform console, navigate to Menu > Services > Compute > Server > Bare Metal Server.
2. Click [Create server].
3. When the server creation interface appears, follow the steps in order:
   • 1. Select server image
   • 2. Configure server settings
   • 3. Configure authentication key
   • 4. Configure network access
   • 5. Review final settings

1. Select server image

Select the server image to create.

1. Select the server type.
   • Bare Metal server (general): General Bare Metal server.
2. From the server image list, click [Next] to the right of the server image you want.
3. In the confirmation popup, review the details and click [OK].

2. Configure server settings

Specify the server configuration information.

1. Select the VPC and Subnet you want to deploy a server to.
   • You can deploy servers only to Subnets intended for Bare Metal server use.
   • To communicate with an external private network, you must deploy the server to a Public Subnet so that it can be assigned a public IP.
   • If you need to create a new VPC or Subnet, click [Create VPC] or [Create Subnet] and proceed with the operation on the page.
2. Select the server type.
   • Select the appropriate number of CPUs and memory capacity for your purpose of use.
3. Select the RAID type (a method of grouping and using multiple disks).
   • RAID 5: RAID level generally used, which is appropriate for typical applications such as web services.
   • RAID 1+0: Appropriate for applications that require high performance and stability, such as DBMS.
4. Set the partition.
   • Number of partitions: Select the number of partitions that logically separate the disk.
   • Partition settings: Set the mount point and capacity for each partition.
     • Mount point: Directory to mount (connect) the disk. It must start with an English letter and only lowercase English letters and numbers can be entered.
     • Partition capacity (GiB): Disk capacity that can be fragmented within the available capacity.
5. Select the pricing plan.
   • Bare Metal servers offer hourly pricing plans only.
   • The fee for the server is charged even when the server is stopped. If the server is not in use, return it.
6. Enter the number of servers and server names.
   • You can create up to 5 servers.
   • If you create multiple servers, you must enter a starting number. You can only enter numbers in the starting number, and "001" is the default value.
   • The server name must be 3-15 characters long and begin with a letter. They must contain lowercase letters, numbers, and/or hyphens. If you do not enter a server name, the name is automatically created.
   • Set the hostname to the server name you entered. Selecting the checkbox applies the hostname set in the server OS as the server name.
7. Enter the network interfaces to assign to the server.
   • You must register the eth0 IP, and it becomes the server's default private IP.
     • Enter the desired IP, and then click [Add]. The IP must be within the selected Subnet range.
     • Leaving the IP field blank and clicking the [Add] automatically allocates it.
   • Apart from the default eth0, you can create 1 additional network interfaces. However, it requires a Private Subnet for Bare Metal servers in the same zone. The entered value becomes the additional private IP for the server.
   Note

   To network interfaces, see the following usage tips related:

   • You can utilize network interfaces to distinguish multiple servers intuitively based on their purpose of use.
     Example: Distinguish 10.1.1.1x as the development server and 10.1.1.2x as the production server.
   • Create multiple network interfaces in the following cases:
     • When you need different Subnets based on the characteristics of applications to be used on servers.
     • When you designed Subnets separated by purpose but need to integrate them into one server for use.
   • You can have up to 2 network interfaces, including 1 additional interface.
8. Select the server return protection status.
   • Return protection helps prevent the accidental return of servers. Enable return protection for servers with essential roles.
9. If necessary, enter a note or select an automatic execution script.
   • Selecting an automatic execution script runs a script automatically after server creation is complete and allows you to configure the initial environment of the server. You must create scripts in the Init Script menu beforehand to select the option.
10. If you have completed all settings, click [Next].

3. Configure authentication key

Select an authentication key you have or create a new one and click [Next].

• To create a new authentication key, select Create new authentication key, enter the authentication key name, and click [Create and save authentication key].

4. Configure network access

Select an ACG that you have or create a new one and click [Next].

• You can select up to 3 ACGs.
• default-acg is the ACG provided by default.
• The setting is available only when creating a new network interface; if you use the existing network interfaces, the preset ACG is applied automatically.

5. Review final settings

Check the details, and then click [Create server].

Manage general server

You can manage a Bare Metal server and change its settings in the same way as for a regular server. For more information, see Manage server.

Create GPU A100 server

To use a Bare Metal server, you must create one from the NAVER Cloud Platform console. To create a Bare Metal server, follow these steps:

1. In the VPC environment of the NAVER Cloud Platform console, navigate to Menu > Services > Compute > Server > Bare Metal Server.
2. Click [Create server].
3. When the server creation interface appears, follow the steps in order:
   • 1. Select a GPU A100 server image
   • 2. Set a GPU A100 server
   • 3. Set a GPU A100 authentication key
   • 4. Set GPU A100 server network access
   • 5. Perform a final review of GPU A100 server creation

1. Select a GPU A100 server image

Select the server image to create.

1. Select the server type.
   • Bare Metal server (GPU A100): GPU A100 server with local NVMe disks.
2. From the server image list, click [Next] to the right of the server image you want.
3. In the confirmation popup, review the details and click [OK].

2. Set a GPU A100 server

Specify the server configuration information.

1. Select the VPC and Subnet you want to deploy a server to.
   • You can deploy servers only to Subnets intended for Bare Metal server use.
   • To communicate with an external private network, you must deploy the server to a Public Subnet so that it can be assigned a public IP.
   • If you need to create a new VPC or Subnet, click [Create VPC] or [Create Subnet] and proceed with the operation on the page.

2. Select the server type.
   • Select the appropriate number of CPUs and memory capacity for your purpose of use.
3. Select the pricing plan.
   • Bare Metal servers offer hourly pricing plans only.
   • The fee for the server is charged even when the server is stopped, so if the server is not in use, return it.
4. Enter the number of servers and server names.
   • You can create up to 5 servers.
   • If you create multiple servers, you must enter a starting number. You can only enter numbers in the starting number, and "001" is the default value.
   • The server name must be 3-15 characters long and begin with a letter. They must contain lowercase letters, numbers, and/or hyphens. If you do not enter a server name, the name is automatically created.
   • Set the hostname to the server name you entered. Selecting the checkbox applies the hostname set in the server OS as the server name.
5. Set the Pool and cluster mode to create a Fabric Cluster for clustering multiple servers.
   • Pool: A logical group of GPU A100 servers. Cluster configuration is only possible within the same Pool.
   • Fabric Cluster: Whether the server is included in a cluster. Select N (Single Mode) to set as single mode, and select Y (Cluster Mode) to set as cluster mode.
6. Enter the network interfaces to assign to the server.
   • You must register the eth0 IP, and it becomes the server's default private IP.
     • Enter the desired IP, and then click [Add]. The IP must be within the selected Subnet range.
     • Leaving the IP field blank and clicking the [Add] automatically allocates it.
   • Apart from the default eth0, you can create 1 additional network interfaces. However, it requires a Private Subnet for Bare Metal servers in the same zone. The entered value becomes the additional private IP for the server.
   Note

   To network interfaces, see the following usage tips related:

   • You can utilize network interfaces to distinguish multiple servers intuitively based on their purpose of use.
     Example: Distinguish 10.1.1.1x as the development server and 10.1.1.2x as the production server.
   • Create multiple network interfaces in the following cases:
     • When you need different Subnets based on the characteristics of applications to be used on servers.
     • When you designed Subnets separated by purpose but need to integrate them into one server for use.
7. Select whether the GPU driver is installed.
   • The GPU driver is not installed by default.
   • To use the default GPU driver provided by the NAVER Cloud Platform, select the corresponding item. If selected, it is automatically installed when the server is created.
8. Select the server return protection status.
   • Return protection helps prevent the accidental return of servers. Enable return protection for servers with essential roles.
9. If necessary, enter a note or select an automatic execution script.
   • Selecting an automatic execution script runs a script automatically after server creation is complete and allows you to configure the initial environment of the server. You must create scripts in the Init Script menu beforehand to select the option.
10. If you have completed all settings, click [Next].

3. Set a GPU A100 server authentication key

Select an authentication key you have or create a new one and click [Next].

• To create a new authentication key, select Create new authentication key, enter the authentication key name, and click [Create and save authentication key].

4. Set GPU A100 server network access

Select an ACG that you have or create a new one and click [Next].

• You can select up to 3 ACGs.
• default-acg is the ACG provided by default.
• The setting is available only when creating a new network interface; if you use the existing network interfaces, the preset ACG is applied automatically.

5. Perform a final review of GPU A100 server creation

Check the details, and then click [Create server].

Manage GPU A100 server

This guide describes how to check the GPU A100 server created on NAVER Cloud Platform.

GPU driver and essential software

Check the GPU driver and essential software on the server. The inspection method is the same as that for KVM A100. For more information, see the following guides:
Detailed check elements are as follows:

• Driver version
• CUDA version
• Docker and NVIDIA Docker
• Required daemons (nvidia-fabricmanager, nv_peer_mem)
• NVLink topology

InfiniBand

InfiniBand is a standard network communication technology used in High-Performance Computing (HPC), offering the advantages of high bandwidth and low latency.
Infiniband is used for communication between GPU A100 servers, so you must check it. The inspection method for InfiniBand is the same as that for KVM A100.
For more information, see the following guides:

• 1. Prerequisites for Infiniband communication between servers
• 2. Check Infiniband connection status
• 3. Communication test

GPU A100 NVMe Local Disk

GPU A100 provides 4 NVMe SSDs of 7.68 TB in amount.
Disks are provided without being partitioned or mounted, so you need to check the basic recognition status of the disk and proceed with partitioning and mounting.
The detailed procedures are as follows:

• 1. Check NVMe disk recognition
• 2. Create and mount file system
• 3. Auto mount using fstab

1. Check NVMe disk recognition

To check if the NVMe disk is recognized normally, follow these steps:

1. Install the nvme-cli package.
2. Run the nvme list command to check the information on the mounted NVMe device.

   # nvme list
   Node             SN                   Model                                    Namespace Usage                      Format           FW Rev
   ---------------- -------------------- ---------------------------------------- --------- -------------------------- ---------------- --------
   /dev/nvme0n1     S4BGNC0R700330       SAMSUNG MZQLB7T6HMLA-00007               1          45.06  kB /   7.68  TB      4 KiB +  0 B   EDB5202Q
   /dev/nvme1n1     S4BGNC0R700328       SAMSUNG MZQLB7T6HMLA-00007               1          45.06  kB /   7.68  TB      4 KiB +  0 B   EDB5202Q
   /dev/nvme2n1     S4BGNW0R702764       SAMSUNG MZQLB7T6HMLA-00007               1          45.06  kB /   7.68  TB      4 KiB +  0 B   EDB5202Q
   /dev/nvme3n1     S4BGNA0R400116       SAMSUNG MZQLB7T6HMLA-00007               1          45.06  kB /   7.68  TB      4 KiB +  0 B   EDB5202Q
   

2. Create and mount file system

Create a file system and mount the disk. You can select and build the form of the file system from the following:

xfs

To create an xfs file system and mount a disk, follow these steps:

1. Run the mkfs.xfs command for formatting into an xfs system for all NVMe devices (/dev/nvme0n1 ~ /dev/nvme3n1).

   mkfs.xfs: /dev/nvme0n1 appears to contain a partition table (gpt).
   mkfs.xfs: Use the -f option to force overwrite.  <--If you encounter the following warning, add the -f (--force) option.
   
   # mkfs.xfs /dev/nvme0n1 -f
   meta-data=/dev/nvme0n1           isize=512    agcount=7, agsize=268435455 blks
               =                       sectsz=4096  attr=2, projid32bit=1
               =                       crc=1        finobt=0, sparse=0
   data     =                       bsize=4096   blocks=1875366486, imaxpct=5
               =                       sunit=0      swidth=0 blks
   naming   =version 2              bsize=4096   ascii-ci=0 ftype=1
   log      =internal log           bsize=4096   blocks=521728, version=2
               =                       sectsz=4096  sunit=1 blks, lazy-count=1
   realtime =none                   extsz=4096   blocks=0, rtextents=0
   
   # blkid | grep /dev/nvme0n1
   /dev/nvme0n1: UUID="1068a11f-df24-48c4-8506-dd522fea6bd2" TYPE="xfs"
   

2. Mount the formatted NVMe disk at the desired mount point.

   # mkdir /mnt/nvme-disk1
   # mount /dev/nvme0n1 /mnt/nvme-disk1
   # df /mnt/nvme-disk1
   Filesystem      1K-blocks  Used  Available Use% Mounted on
   /dev/nvme0n1   7499379032 33220 7499345812   1% /mnt/nvme-disk1
   

ext4

To create an ext4 file system and mount the disk, follow these steps:

1. Run the mkfs.ext4 command for formatting into an ext4 system for all NVMe devices (/dev/nvme0n1 ~ /dev/nvme3n1).

   # mkfs.ext4 /dev/nvme1n1
   mke2fs 1.42.9 (28-Dec-2013)
   Discarding device blocks: done
   Filesystem label=
   OS type: Linux
   Block size=4096 (log=2)
   Fragment size=4096 (log=2)
   Stride=0 blocks, Stripe width=0 blocks
   234422272 inodes, 1875366486 blocks
   93768324 blocks (5.00%) reserved for the super user
   First data block=0
   Maximum filesystem blocks=4024434688
   57232 block groups
   32768 blocks per group, 32768 fragments per group
   4096 inodes per group
   Superblock backups stored on blocks:
           32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
           4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968,
           102400000, 214990848, 512000000, 550731776, 644972544
   
   Allocating group tables: done
   Writing inode tables: done
   Creating journal (32768 blocks): done
   Writing superblocks and filesystem accounting information: done
   
   # blkid | grep /dev/nvme1n1
   /dev/nvme1n1: UUID="fcc67093-6f37-4456-b5cf-70f2a6157495" TYPE="ext4"
   

2. Mount the formatted NVMe disk at the desired mount point.

   # mkdir /mnt/nvme-disk2
   # mount -t ext4 /dev/nvme1n1 /mnt/nvme-disk2
   # df /mnt/nvme-disk2
   Filesystem      1K-blocks  Used  Available Use% Mounted on
   /dev/nvme1n1   7442192600 90136 7067012784   1% /mnt/nvme-disk2
   

3. Auto mount using fstab

After formatting, the disk mounted using the mount command must be mounted again each time the operating system reboots. Set up fstab to mount the disk automatically upon boot.

To set up fstab for auto mounting, follow these steps:

1. Run the blkid command and check UUID by partition.

   # blkid | grep '/dev/nvme'
   /dev/nvme0n1: UUID="1068a11f-df24-48c4-8506-dd522fea6bd2" TYPE="xfs"
   /dev/nvme1n1: UUID="fcc67093-6f37-4456-b5cf-70f2a6157495" TYPE="ext4"
   /dev/nvme2n1: UUID="844741dc-306c-44a2-a18c-9b1c780d36df" TYPE="xfs"
   /dev/nvme3n1: UUID="e101556f-69c3-4dc7-870a-0e037a97c49f" TYPE="xfs"
   

2. Open the /etc/fstab file to add mount information and save it.
   • CentOS

   # cat /etc/fstab 
   
   #
   # /etc/fstab
   # Created by anaconda on Thu Nov 24 12:05:24 2022
   #
   # Accessible filesystems, by reference, are maintained under '/dev/disk'
   # See man pages fstab(5), findfs(8), mount(8) and/or blkid(8) for more info
   #
   UUID=8518300e-9c84-470c-9d1f-4f6cd625752f /                       xfs     defaults        0 0
   UUID=240240cc-79d5-43af-a3e0-ae06070c682a /boot                   xfs     defaults        0 0
   UUID=E036-A50F          /boot/efi               vfat    defaults,uid=0,gid=0,umask=0077,shortname=winnt 0 0
   UUID=9505f0f2-3e43-4093-b627-ac62508f4ba0 swap                    swap    defaults        0 0
   
   <!-- Mount information entry format: UUID={UUID of the partition}  {mount point}  {file system type}  {mount option} -->
   UUID=1068a11f-df24-48c4-8506-dd522fea6bd2 /mnt/nvme-disk1         xfs     defaults        0 0
   UUID=fcc67093-6f37-4456-b5cf-70f2a6157495 /mnt/nvme-disk2         ext4    defaults        0 0
   UUID=844741dc-306c-44a2-a18c-9b1c780d36df /mnt/nvme-disk3         xfs     defaults        0 0
   UUID=e101556f-69c3-4dc7-870a-0e037a97c49f /mnt/nvme-disk4         xfs     defaults        0 0
   

   • Ubuntu

   # /etc/fstab: static file system information.
   #
   # Use 'blkid' to print the universally unique identifier for a
   # device; this may be used with UUID= as a more robust way to name devices
   # that works even if disks are added and removed. See fstab(5).
   #
   # <file system> <mount point>   <type>  <options>       <dump>  <pass>
   # / was on /dev/sda3 during curtin installation
   /dev/disk/by-uuid/4723f105-37b0-4602-8cc2-173419cdd0d8 / xfs defaults 0 1
   # /boot was on /dev/sda2 during curtin installation
   /dev/disk/by-uuid/c9216d88-ddd1-4aac-a4d5-da645234a7bb /boot ext4 defaults 0 1
   # /boot/efi was on /dev/sda1 during curtin installation
   /dev/disk/by-uuid/2753-311A /boot/efi vfat defaults 0 1
   /swap.img       none    swap    sw      0       0
   
   <!-- Mount information entry format: /dev/disk/by-uuid/{UUID of the partition}  {mount point}  {file system type}  {mount option} -->
   /dev/disk/by-uuid/1068a11f-df24-48c4-8506-dd522fea6bd2 /mnt/nvme-disk1         xfs     defaults        0 1
   /dev/disk/by-uuid/fcc67093-6f37-4456-b5cf-70f2a6157495 /mnt/nvme-disk2         ext4    defaults        0 1
   /dev/disk/by-uuid/844741dc-306c-44a2-a18c-9b1c780d36df /mnt/nvme-disk3         xfs     defaults        0 1
   /dev/disk/by-uuid/e101556f-69c3-4dc7-870a-0e037a97c49f /mnt/nvme-disk4         xfs     defaults        0 1