← Solutions Notes · กลับหน้าแรก

Huawei · Communication matrix

OceanProtect X series backup appliance

Every network connection the appliance needs in the FusionCompute (LAN-based) backup scenario — 25 flows drawn by the plane they run on, then listed in full with protocol, authentication and encryption.

Scenario
Scenario
FusionCompute (LAN-based)
Flows
25
Listening ports
21 distinct
Applies to
OceanProtect DataBackup 1.3.0 and later
EXTERNAL INFRASTRUCTURE SERVICES — APPLIANCE INITIATES MANAGEMENT & O&M CLIENTS NTP server external time source Third-party CA server certificate service External CA server certificate service AD / Kerberos server domain authentication Remote OceanProtect replication · DR · archive target 123/TCP-UDP time synchronization 8443/TCP one-click cert update 1024–65535 automatic cert update 88/TCP · 464/TCP-UDP Kerberos & AD auth 8088/TCP remote appliance link OceanProtect X series BACKUP APPLIANCE Storage controller management + backup network planes Management browser DeviceManager / WebUI Third-party NMS SmartKit · DME Admin host SSH / CLI 8088/TCP · 25080/TCP DeviceManager & WebUI over TLS 25081/TCP REST API — SmartKit, DME, NMS 22/TCP · SSH CLI login · remote power-on PROTECTED ENVIRONMENT · FUSIONCOMPUTE PLATFORM VM ON THE PLATFORM Backup client host ProtectAgent + NFS client CentOS · Red Hat · SUSE FusionCompute VRM virtualization platform 30065/TCP · 25082/TCP signaling · registration · heartbeat 59520–59559/TCP · 22/TCP job delivery · agent installation 2049 · 2050 · 2051 · 2052 · 111 53/UDP · 762/2053–2059/UDP NFS data, mount, locks, RPC, DNS, quota 22/TCP · 138 · 445/TCP file-level restore to the target VM (SSH / SMB) 7443/TCP · TLS VM backup & restore control
Arrows point from the side that opens the connection to the side that listens; each label carries the listening port and what it is for. The appliance only ever initiates outward to infrastructure services and the remote appliance (top), to the ProtectAgent for job delivery and installation, and to the target VM during a file-level restore — everything else is opened towards it. The ProtectAgent host and the NFS client are one and the same machine in this scenario, so they are drawn as a single box — and that host itself runs as a VM on the platform, inside the dashed boundary. The file-level restore path is drawn to that same box; its real destination is whichever VM on the platform is being restored into. Switching the scenario above changes only one connection — the one from that host to the platform it protects.

What the client asks the platform for

A zoom-in on the single API line in the map above.

OPENSTACK PLATFORM INSTANCE ON THE PLATFORM Backup client instance ProtectAgent Keystone identity Nova compute Cinder block storage Neutron networking 1 · authenticate token used on every call below 2 · instance info attach and detach the temp volume 3 · snapshot volumes create and delete the temp volume 4 · network config read when restoring an instance PROTECTED INSTANCES Instances being backed up never touched by the agent directly Their volumes the snapshot source acts on snapshots All four calls go to the same endpoint — the 443/TCP line in the map above.
Keystone is only the front door: the client authenticates there once, then spends that token on Nova, Cinder and Neutron. The backup data itself never crosses this path — Cinder snapshots the protected volumes, and the copy is read from a temporary volume attached to the client.

How a backup actually runs

Six steps, from the OpenStack data protection guide (V200R001C33).

Six steps as documented for OpenStack (V200R001C33) — confirm the FusionCompute wording before quoting it.

Backup service process: the OceanProtect appliance creates a file system and starts the job; the client snapshots the cloud disks through the OpenStack API and creates a snapshot volume; the appliance reads that volume into the file system and takes a file system snapshot; the client then removes the volume and snapshots.
Backup service process, as drawn in the OpenStack data protection guide (V200R001C33). The numbers match the steps below.
  1. 1
    OceanProtect

    The data protection engine creates the target file system on the OceanProtect and tells the client to start the job.

  2. 2
    Client

    The client asks Cinder to snapshot the instance’s volumes, then create a new volume from those snapshots — this is the 443 API path, authenticated through Keystone and calling Nova and Cinder.

  3. 3
    Client

    That temporary volume is attached to the client instance and mounted there, so the data is read without touching the running production instance.

  4. 4
    OceanProtect

    The OceanProtect reads the attached volume and writes it into the file system over the backup path. Data is read in fixed-size blocks and hashed; on an incremental backup a block is written only when its hash differs from the one already recorded.

  5. 5
    OceanProtect

    When the backup completes, the OceanProtect takes a file system snapshot. That snapshot is the restore point.

  6. 6
    Client

    The client unmounts and detaches the temporary volume, then deletes it along with the volume snapshots it created in step 2.

Steps 2 and 6 run over the platform API connection; steps 3 to 5 run over the backup path between the client host and the appliance. That split is why both sets of ports have to be open.

Saying it the way the platform team says it

The guide uses cloud-vendor wording; these are the words an OpenStack operator uses.

In the product documentationOn the platformWhy it matters
Cloud server / cloud hostInstanceA Nova server. “Cloud host” is not a word the platform team uses.
Cloud diskVolumeA Cinder volume. Every disk-level statement in the guide is really about volumes.
Cloud disk snapshotVolume snapshotA Cinder snapshot, which the backup then turns into a temporary volume.
Local diskEphemeral diskNot backed up — worth stating plainly, because instances booted from ephemeral storage are out of scope.
Shared diskMulti-attach volumeAlso out of scope.
Bare metal server (BMS)Bare metal node (Ironic)Cannot be backed up at all.
System diskRoot volumeFile-level restore needs it in the copy.
Keystone V3 addressPublic identity v3 endpointWhat openstack endpoint list returns for the public interface.
Domain, projectSame wordsKeystone domain and project carry their usual meaning here.

Before you register OpenStack

Getting the Keystone V3 address on the community version.

  1. Log in to the management page of the OpenStack community version.
  2. In the navigation tree on the left, choose Admin › System › System Information.
  3. On the Services tab, read the value of Public for the keystone service.

Version differences aside: if the Public value does not end with /v3, add /v3 to the end of it to get the Keystone V3 address. If it already ends with /v3, the Public value is the Keystone V3 address.

OpenStack Horizon, Admin, System, System Information page. On the Services tab the keystone row shows its identity service endpoints, with the Public endpoint highlighted.
The Public endpoint of the keystone row on the Services tab is the address to take.

Client host prerequisites

Checks to run on the Linux host before the ProtectAgent is installed.

Where it goes: deploy the client as an instance in the same Keystone domain as the instances to be backed up or restored.

RequirementWhyDependency packageAutomatic push installManual install
Port 22 with sshd and sftpAutomatic push installation connects over port 22 and uses the sshd and sftp services. Check the host before installing.NoManually configuredN/A
Nginx port open in the host firewallThe Nginx listening port carries traffic between the OceanProtect and the client host. With the host firewall on, allow input and output on that port.NoManually configuredManually configured
NFS installedNeeded when the NFS protocol carries data between the client host and the OceanProtect.YesConfigured on the GUIManually configured
Haveged or rng-toolsBelow 100 OS entropy, true random numbers come slowly. Haveged is advised on a VM, a cloud host, an Arm server or any server with no hardware random number generator; rng-tools on the rest.YesConfigured on the GUIManually configured
Command line toolsperl and one of awk / gawk / nawk, plus the net-tools and iproute packages.YesConfigured on the GUIManually configured
Compression toolsAutomatic push: tar 1.22 or later. Manual install: unzip for a ZIP package, tar 1.22 or later for a TAR package.YesConfigured on the GUIManually configured

Full port list

PlaneSourceDestinationPort (listening)ProtocolAuthenticationEncryptionPurposeFeature scenario
Management & O&MDeviceManager Web clientStorage controller8088TCPUsername and passwordTLSDeviceManager web service for system management and maintenance.System management - basics
Management & O&MWeb clientStorage controller25080TCPUsername and passwordTLSOceanProtect WebUI on the management network plane.System management - basics
Management & O&MExternal system (SmartKit / DME / NMS)Storage controller25081TCPUsername and passwordTLSRESTful API invoked by third-party network management systems.System management - basics
Management & O&MAny hostStorage controller22TCPUsername and passwordSSHSSH login to the storage CLI for O&M channel management.System management - basics
Management & O&MAny hostStorage controller22TCPUsername and passwordSSHRemote power-on after the system has been properly powered off.System management - basics
Management & O&MAny hostDisk enclosure22TCPUsername and passwordSSHSSH login to the disk enclosure CLI.System management - basics
Backup & recovery pathClient host (independent deployment)FusionCompute VRM node7443 TCPUsername and passwordTLSClient connects to the FusionCompute platform to back up and restore VMs.Basic protection service - basics
Backup & recovery pathClient host (independent deployment)OpenStack controller node — Keystone V3 address443 (default) · per-service (community edition) TCPUsername and passwordTLSClient invokes the OpenStack service APIs for interconnection, backup and restoration.Basic protection service - basics
Backup & recovery pathClient hostStorage controller30065TCPCertificate authenticationTLSCarries backup and restoration service signaling from the ProtectAgent.Basic protection service - basics
Backup & recovery pathClient hostStorage controller25082TCPCertificate authenticationTLSProtectAgent registration, heartbeat and status reporting on the service plane.System management - basics
Backup & recovery pathStorage controllerClient host59520–59559TCPCertificate authenticationTLSAppliance delivers backup and restoration jobs to the external ProtectAgent.System management - basics
Backup & recovery pathStorage controllerLinux and other UNIX hosts / client host22TCPUsername and passwordNoneManagement channel used to install the ProtectAgent client.System management - basics
Backup & recovery pathStorage controllerTarget VM for restoration22 (Linux SSH) · 138, 445 (Windows SMB)TCPUsername and passwordNoneFile-level restoration from a copy to the selected target VM.Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller2049TCPNIS / LDAP / username and passwordNoneNFS read and write operations — live mount and instant recovery.Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller2050TCP/UDPNIS / LDAP / username and passwordNoneMount port for NFS shares (service network ports only).Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller2051TCP/UDPNIS / LDAP / username and passwordNoneNSM — client instructs the server to clear lock resources.Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller2052TCP/UDPNIS / LDAP / username and passwordNoneNLM — file lock and unlock.Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller111TCP/UDPNoneNoneRPC registration service lookup (portmapper).Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller53UDPNoneNoneBuilt-in DNS load balancing for the NAS service IP. Disabled by default.Basic protection service - basics
Backup & recovery pathNFS client / client hostStorage controller762 / 2053–2059UDPNoneNoneRquota — queries and manages NFS service quotas.Basic protection service - basics
External servicesStorage controllerNTP server (external)123TCP/UDPNoneNoneTime synchronization once an NTP server is set on the management plane.System management - basics
External servicesStorage controllerThird-party CA server8443 TCPCertificate authenticationTLSOne-click certificate update against a third-party CA.System management - basics
External servicesStorage controllerExternal CA server1024–65535HTTP/TCPCertificateTLSAutomatic certificate update.System management - basics
External servicesStorage controllerAD domain server (external)88TCPNTLM / KerberosNoneKerberos authentication once a Kerberos server is set on the service network.Basic protection service - backup channel encryption
External servicesStorage controllerAD domain server (external)464TCP/UDPNTLM / KerberosNoneAD domain authentication.Basic protection service - backup channel encryption
External servicesStorage controllerExternal storage — remote OceanProtect appliance8088TCPUsername and passwordTLSConnects an external storage unit (a second OceanProtect appliance or other target) for backup, restoration, archiving and replication.System management - basics

Limits worth checking before you commit

From the OpenStack data protection guide (V200R001C33).

From the OpenStack data protection guide (V200R001C33) — the FusionCompute figures may differ.

What can be protected

  • Instances only — bare metal nodes (Ironic) cannot be backed up.
  • Cinder volumes only — ephemeral (local) disks and multi-attach (shared) volumes cannot be backed up.
  • The instance must be ACTIVE, SHUTOFF or SUSPENDED.
  • Crash-consistent backup only. Application-consistent backup is not supported.
  • Backup granularity is the whole instance or selected volumes; restore is at instance, volume or file level.

Space and job behaviour

  • A backup creates a temporary volume, so free space in the Cinder backend during the job must exceed the size of the volume being backed up.
  • Snapshot-consistent backup and linked-clone volumes need the instance’s volumes on one OceanStor Dorado 6.1.6SPH6 or later, with FusionSphere OpenStack 24.1 or later.
  • One backup job per resource at a time, and backup and restore cannot run together on the same resource.
  • After a restore to the original location, the next backup is a full backup.
  • Before restoring, free space at the target must exceed the size of the copy.

Indexing and file-level restore

  • Index cap per copy: 20 million files on X3000/X6000, 50 million on X8000/X8100/X9000/X9100, 30 million on E6000.
  • At most 200,000 files per restore job.
  • Linux ext2/3/4 and XFS; Windows NTFS, FAT and FAT32. ReFS, swap partitions and encrypted disks cannot be parsed.
  • The copy must include the root volume, and only a single-OS copy can be parsed.
  • Mount information must be in fstab; VIRTIO disks parse only when it is written as a UUID.
  • File names must be UTF-8 and shorter than 256 bytes.
  • Supported guests: Linux, Windows 10 or earlier, Windows Server 2022 or earlier.