Skip to content

Deployment Requirements

Before deploying DCE 5.0, software planning, hardware planning, and network planning must be completed in advance.

Operating System Requirements

Architecture OS Tested OS / Kernel Notes (Installation Guide)
AMD 64 Red Hat 8.X Red Hat 8.4
4.18.0-305.el8.x86_64		 Offline Install DCE 5.0 Enterprise
Red Hat 7.X Red Hat 7.9
3.10.0-1160.el7.x86		 Offline Install DCE 5.0 Enterprise
Red Hat 9.X Red Hat 9.2
5.14.0-284.11.1.el9_2.x86_64		 Offline Install DCE 5.0 Enterprise
Ubuntu 22.04 5.15.0-78-generic Offline Install DCE 5.0 Enterprise
Ubuntu 24.04 / Offline Install DCE 5.0 Enterprise
Rocky Linux 9.2 5.14.0-284.11.1.el9_2.x86_64 Offline Install DCE 5.0 Enterprise
UnionTech UOS V20 (1020a) 5.4.0-125-generic Deploy DCE 5.0 Enterprise on UOS V20 (1020a)
openEuler 22.03 5.10.0-60.18.0.50.oe2203.x86_64 Offline Install DCE 5.0 Enterprise
Oracle Linux R9/R8 U1 5.15.0-3.60.5.1.el9uek.x86_64 Deploy DCE 5.0 Enterprise on Oracle Linux R9 U1
TencentOS Server 3.1 5.4.119-19.0009.14 Deploy DCE 5.0 Enterprise on TencentOS Server 3.1
ARM 64 Kylin OS V10 SP2 4.19.90-24.4.v2101.ky10.aarch64 Offline Install DCE 5.0 Enterprise
Kylin OS V10 SP3 4.19.90-89.11.v2401.ky10.aarch64 Offline Install DCE 5.0 Enterprise

Note

Kernel Requirements

Some components or features have specific kernel version requirements. Before deployment, select the kernel version according to the following table:

Component/Feature Kernel Version
Container management GPU capability ≥ 5.15
Cilium ≥ 5.12
HwameiStor DRBD capability DRBD-supported kernel versions
KubeVirt > 4.11
Merbridge ≥ 5.7

⚠️ Kernel Notes

  1. When the kernel version is lower than 5.9 and kube-proxy uses ipvs mode, accessing in-cluster services via Service may occasionally experience a 1-second delay or Service access failures after backend upgrades. For details, see Kubernetes community Issue #81775. The following workarounds are available:

    • Upgrade the kernel to 5.9 or later
    • Switch the kube-proxy mode to iptables
    • Update kernel parameters:
      net.ipv4.vs.conntrack=1 + net.ipv4.vs.conn_reuse_mode=0 + net.ipv4.vs.expire_nodest_conn=1

  2. Automatic Ubuntu kernel updates may cause unexpected system reboots. If your software depends on a specific kernel version, automatic upgrades to a newer kernel may introduce compatibility issues.

Hardware Requirements

CPU, Memory, and Disk

Type Requirement
CPU No overcommit
Memory No overcommit
Disk IOPS > 500, throughput > 200 MB/s

CPU, Memory, and Disk Requirements in Beginner Mode

Refer to the Beginner Mode Description.

Count Server Role Purpose CPU Memory System Disk Unpartitioned Disk
1 All-in-one Image registry, chart museum, and the global cluster itself 24 cores 48 GB 300 GB 400 GB

If disk partitioning is required, allocate disk resources according to the table below:

Server Role Disk Partition / Directory Recommended Free Space Notes
All-in-one /var/lib >50 GB On master nodes, this path stores local images, etcd, monitoring, logs, etc., and requires reserved buffer space. Worker nodes use less space, but a unified standard is recommended.
All-in-one / (root filesystem) >70 GB If /home and /var/lib are not separate partitions, their required space must be added to the root partition.
All-in-one /home >150 GB /

4-Node Mode CPU, Memory, and Disk Requirements

Refer to the 4-Node Mode Description.

Count Server Role Purpose CPU Memory System Disk Unpartitioned Disk
1 Bootstrap node 1. Run the installation program
2. Run the required image registry and chart museum		 2 cores 4 GB 200 GB -
3 Control plane 1. Run DCE 5.0 components
2. Run Kubernetes system components		 16 cores 32 GB 100 GB 200 GB

If disk partitioning is required, allocate disk resources according to the table below:

Server Role Disk Partition / Directory Recommended Free Space Notes
Bootstrap node /var/lib >30 GB Mainly used for bootstrap Podman container storage (/var/lib/containers)
Bootstrap node / (root filesystem) >150 GB Stores offline installation resources (ISO, Addons, Binaries, Images, etc.)
Bootstrap node /home >50 GB If /home and /var/lib are not separate partitions, their required space must be added to the root partition.
Bootstrap node (upgrade considered) /var/lib >30 GB /
Bootstrap node (upgrade considered) / (root filesystem) >50 GB /
Bootstrap node (upgrade considered) /home >300 GB Space requirements double if old installation packages are retained for upgrades
Control plane /var/lib >50 GB /
Control plane / (root filesystem) >50 GB /

Refer to the 7-Node Mode Description.

Count Server Role Purpose CPU Memory System Disk Unpartitioned Disk
1 Bootstrap node 1. Run the installation program
2. Run the required image registry and chart museum		 2 cores 4 GB 200 GB -
3 Master 1. Run DCE 5.0 components
2. Run Kubernetes system components		 8 cores 16 GB 100 GB 200 GB
3 Worker Run logging-related components separately 8 cores 16 GB 100 GB 200 GB

If disk partitioning is required, allocate disk resources according to the table below:

Server Role Disk Partition / Directory Recommended Free Space Notes
Bootstrap node /var/lib >30 GB Mainly used for bootstrap Podman container storage (/var/lib/containers)
Bootstrap node / (root filesystem) >150 GB Stores offline installation resources (ISO, Addons, Binaries, Images, etc.)
Bootstrap node /home >50 GB If /home and /var/lib are not separate partitions, their required space must be added to the root partition.
Bootstrap node (upgrade considered) /var/lib >30 GB /
Bootstrap node (upgrade considered) / (root filesystem) >50 GB /
Bootstrap node (upgrade considered) /home >300 GB Space requirements double if old installation packages are retained for upgrades
Master /var/lib >50 GB /
Master / (root filesystem) >50 GB /
Worker /var/lib >30 GB /
Worker / (root filesystem) >50 GB /

etcd Disk Recommendations

Since etcd writes data to disk and persists it on disk, its performance depends on disk performance. etcd is extremely sensitive to write latency, and latency issues may cause etcd to lose heartbeats.

Recommendations:

  • Run etcd on machines backed by low-latency and high-throughput SSD or NVMe disks whenever possible.
  • Use SSDs as the minimum choice; NVMe drives are preferred in production environments.

Network Requirements

Network Topology

Assuming VIP is used as the load balancing method for the global cluster:

Network-Topology

Network Requirements

Resource Requirement Description
istioGatewayVip 1 If the load balancing mode is MetalLB, a VIP must be specified to provide access to the DCE UI and OpenAPI.
insightVip 1 If the load balancing mode is MetalLB, a VIP must be specified as the Insight data collection endpoint for the global cluster. Subcluster insight-agents report data to this VIP.
Network speed 1000 Mb/s At least 1 Gbps; 10 Gbps recommended
Protocol - IPv6 supported
Reserved IP ranges Two ranges required For Pods (default 10.233.64.0/18) and Services (default 10.233.0.0/18). If already in use, customize other ranges to avoid IP conflicts.
Routing - Servers must have a default route or a route pointing to 0.0.0.0.
NTP server addresses 1–4 Ensure accessible NTP server IP addresses are available in your data center.
DNS server addresses 1–2 If your applications require DNS services, prepare accessible DNS server IP addresses.

Client Browser Requirements

  • Firefox ≥ v49
  • Chrome ≥ v54

Comments