VMware ESXi

Configuring the Load Balancing Mode

Perform the following operations to configure the load balancing mode:

1. On DeviceManager, choose Services > Hosts. Select the desired host, click on the right, and choose Modify.
   Figure 7-16 Modifying the host properties
   
   • The information displayed on the GUI may vary slightly with the product version.
   • On DeviceManager of OceanStor Dorado V6 6.0.1 and later versions, the icon is changed to More.

2. On the Modify Host page, set Host Access Mode to Load balancing.
   Figure 7-17 Settings on the local storage system
   

3. Repeat the preceding steps to set Host Access Mode of the remote storage system to Load balancing.
   Figure 7-18 Settings on the remote storage system
   

Configuring the Local Preferred Mode

Perform the following operations to configure the local preferred mode:

1. On DeviceManager, choose Services > Hosts. Select the desired host, click on the right, and choose Modify, as shown in.
   Figure 7-19 Modifying the host properties
   

2. For the local storage system, set Host Access Mode to Asymmetric and Preferred Path for HyperMetro to Yes.
   Figure 7-20 Settings on the local storage system
   

3. For the remote storage system, set Host Access Mode to Asymmetric and Preferred Path for HyperMetro to No.
   Figure 7-21 Settings on the remote storage system
   

For details about the VMware ESXi versions, see the Huawei Storage Interoperability Navigator.

Context

This configuration must be performed separately on all hosts.

UltraPath 21.3.0 and earlier versions support Secure Boot on servers.

Prerequisites

VMware has the following requirements on storage LUN IDs:

• All path LUN IDs of a LUN are consistent for a host.
• All host LUN IDs of a LUN shared by multiple hosts are consistent across these hosts. For details, see Setting LUN Allocations at the VMware office site.

If LUN IDs do not meet the preceding requirements, correct them by referring to "ID Description" in the OceanStor Dorado Host Connectivity Guide for VMware ESXi.

Configuring UltraPath

1. Set the HyperMetro working mode.

   Use either of the following methods to set the HyperMetro working mode for UltraPath:

   Method 1: Run the esxcli upadm set hypermetro workingmode -m auto command to configure UltraPath to automatically adapt the HyperMetro working mode. This setting enables UltraPath to periodically query the host access mode configured on HyperMetro storage systems and adapt its HyperMetro working mode according to the host access mode.

   Method 2: Run the following command to set UltraPath to work in a fixed HyperMetro working mode:

   Table 7-8 Command for setting the HyperMetro working mode for UltraPath

   Command	Example
   set hypermetro workingmode -m [priority | balance] -p primary_array_id	esxcli upadm set hypermetro workingmode -m priority -p 0
   NOTE: In VMware vSphere, adding esxcli upadm in front of a command navigates the user to the UltraPath CLI.

   Table 7-9 describes the command parameters.

   Table 7-9 Parameter description

   Parameter	Description	Default Value
   -m mode	HyperMetro working mode. priority: local preferred mode balance: load balancing mode	priority priority is recommended. balance is applicable when two active-active data centers are in the same building.
   -p primary_array_id	ID of the preferred storage array. The ID is allocated by UltraPath. The storage array that is in the same data center as the application hosts reside is preferred. Run the esxcli upadm show diskarray command to obtain the storage array ID. NOTE: In priority mode, this parameter indicates the storage array to which I/Os are preferentially delivered. In balance mode, this parameter indicates the storage array where the first slice section resides.	None NOTE: Mapping relationship between application hosts and storage arrays: Storage array A is the preferred array for all application hosts in data center A. Storage array B is the preferred array for all application hosts in data center B.

   If you set UltraPath to automatically adapt the HyperMetro working mode, ensure that the host access mode on the storage system is consistent with that on the physical network.

2. Configure the load balancing policy.

   If HyperMetro works in load balancing mode, you can run the esxcli upadm set hypermetro loadbalancemode –m [split-size | round-robin] command to configure the load balancing policy. The following table describes the parameters in the command.

   Parameter	Description	Default Value
   -m mode	Load balancing policy for HyperMetro systems. split-size: slicing mode across storage systems. In this mode, UltraPath delivers I/Os to a specific storage system based on the start addresses of I/Os, slice size, and preferred storage system. For example, if the slice size is 128 MB, the I/Os whose start addresses range from 0 to 128 MB (excluding 128 MB) are preferentially delivered to the preferred storage system and the I/Os whose start addresses range from 128 MB to 256 MB (excluding 256 MB) are delivered to the non-preferred storage system. The default slice size is 128 MB. You can run the esxcli upadm set hypermetro split_size command to change it. round-robin: round-robin mode across storage systems. In this mode, UltraPath selects two storage systems in turn to deliver I/Os.	split-size

3. Set timeout parameters.

   FC networking does not require setting of timeout parameters.

   For iSCSI networking, perform the following operations on ESXi hosts. You must restart the host for the configuration to take effect. Exercise caution when performing this operation.

   1. Obtain the iSCSI adapter name.

      [root@esxi16113:~] esxcfg-scsidevs -a | grep -i iscsi
      vmhba35 iscsi_vmk         online    iqn.XXXX-XX.com.vmware:esxi16113-xxxxxxxxxSCSI Software Adapte

      In the command output, the iSCSI adapter name is vmhba35.

   2. Query the current timeout parameter settings.

      [root@esxi16113:~] esxcli iscsi adapter param get -A vmhba35 | egrep 'NoopOutInterval|NoopOutTimeout|RecoveryTimeout'
      NoopOutInterval       15          15          1    60            true    false
      NoopOutTimeout        10          10          10   30            true    false
      RecoveryTimeout       10          10          1    120           true    false

   3. Set the timeout parameters.

      esxcli iscsi adapter param set -A vmhba35 -k NoopOutInterval -v 1 
      esxcli iscsi adapter param set -A vmhba35 -k NoopOutTimeout -v 10 
      esxcli iscsi adapter param set -A vmhba35 -k RecoveryTimeout -v 1

      • All the preceding commands are available only in VMware ESXi 5.5, 6.0, 6.5, and 6.7. For details on the VMware versions supported in HyperMetro, see the Huawei Storage Interoperability Navigator.
      • The information in bold is the iSCSI adapter. Change it based on the site requirements. (Run the esxcfg-scsidevs -a command to query the iSCSI adapter.)
      • You must restart the host for the configuration to take effect.
      • The settings shorten the path switchover time to about 16s. In comparison, the default ESXi settings may result in an up-to-25s path switchover time.
   4. Verify the modification.

      [root@esxi16113:~] esxcli iscsi adapter param get -A vmhba35 | egrep 'NoopOutInterval|NoopOutTimeout|RecoveryTimeout'
      NoopOutInterval       1           15          1    60            true    false
      NoopOutTimeout        10          10          10   30            true    false
      RecoveryTimeout       1           10          1    120           true    false

Configuring a VMware Cluster

Mandatory configuration items:

• A VM service network requires L2 interworking between data centers so that VM migration between data centers will not affect VM services.
• For VMware vSphere 5.0 u1, later 5.0 versions, and vSphere 5.1:
  • Cluster parameter configuration: Use the vSphere Client to log in to each ESXi host. In the advanced settings, set das.maskCleanShutdownEnabled = True.
  • Host parameter configuration: Log in to each ESXi host using SSH and add Disk.terminateVMOnPDLDefault = True to the /etc/vmware/settings file. After configuring host parameters, restart the host for the configuration to take effect.
• For VMware vSphere 5.5.*, 6.0 u1, and versions between them:
  • Cluster parameter configuration: Use vSphere Web Client to connect to vCenter, go to the cluster HA configuration page, and select Turn on vSphere HA. After configuring cluster parameters, re-enable HA for the configuration to take effect.
  • Host parameter configuration: Log in to each ESXi host using vSphere Client or vCenter and complete the following advanced settings.
    Set VMkernel.Boot.terminateVMOnPDL = True. The parameter forcibly powers off VMs on a datastore when the datastore enters the PDL state.

    Figure 7-22 Boot parameter settings
    
    Set Disk.AutoremoveOnPDL = 0. This setting ensures that datastores in the PDL state will not be automatically removed.

    Figure 7-23 Disk parameter settings
    
• For VMware vSphere 6.0 u2 and later updates:

  After connecting to vCenter through the vSphere Web Client, enter the cluster HA configuration and set the parameters as follows.

  Figure 7-24 vSphere 6.0 cluster configuration
  

• For VMware vSphere 6.5:
  After connecting to vCenter through the vSphere Web Client, enter the cluster HA configuration and set the parameters as follows.

  Figure 7-25 vSphere 6.5 cluster configuration-1
  
  Figure 7-26 vSphere 6.5 cluster configuration-2
  

• For VMware vSphere 6.7 and later updates:

  After connecting to vCenter through the vSphere Web Client, enter the cluster HA configuration and set the parameters as follows.

  Figure 7-27 vSphere 6.7 cluster configuration-1
  
  Figure 7-28 vSphere 6.7 cluster configuration-2
  

Recommended configuration items:

• Configure the vMotion, service, and management networks with different VLAN IDs to prevent network interference.
• Configure the management network to include the vCenter Server management node and ESXi hosts. Deny access from external applications.
• Divide the service network into VLANs to ensure logical isolation and control broadcast domains.
• Configure a DRS group to ensure that VMs can be recovered first in the local data center in the event of the breakdown of a single host.