All Products
Search
Document Center

Server Load Balancer:Modify an NLB instance

Last Updated:Jun 18, 2026

You can modify an NLB instance by updating its availability zones, changing the status of an availability zone, or adjusting the maximum public bandwidth.

Update availability zones

The following table describes the limitations, effective time, billing impact, and use cases for modifying availability zones and subnets.

Limitations

Effective time

Billing impact

Use cases

You cannot reduce the number of availability zones for an NLB instance if it has two or fewer availability zones.

Restrictions on editing availability zones for public-facing NLB instances:

  • The EIPs allocated to a single NLB instance in different availability zones must be of the same type.

  • You can only bind pay-as-you-go (pay-by-data-transfer) EIPs that are not added to an Internet Shared Bandwidth.

Note
  • Before you associate an EIP, make sure that the EIP is not added to an Internet Shared Bandwidth. If you need to add the EIP to an Internet Shared Bandwidth, you can do so in the Load Balancer console after the EIP is associated with the NLB instance.

  • Binding EIPs that are billed by subscription or pay-as-you-go (pay-by-bandwidth) is not supported.

  • A system-allocated EIP is also released when the NLB instance is released. An existing EIP that you associate is retained when the NLB instance is released.

Changes to availability zones usually take effect immediately. However, you may experience a delay of a few minutes due to network latency.

Modifying the availability zone of a public NLB instance may change the associated EIP. EIPs are a paid service, and the actual fees are subject to your bill. For more information, see EIP Billing Overview.

You can edit the availability zone of an instance if its current availability zone or the EIP that is bound to the public NLB instance no longer meets your business requirements.

  1. Log on to the NLB console.

  2. In the top navigation bar, select the region where the NLB instance is deployed.

  3. On the Instances page, find the NLB instance that you want to manage and use one of the following methods to modify its availability zones and subnets:

    • In the Actions column, choose More actions > Modify Zone/Subnet.

    • Click the instance ID. On the Instance Details tab, click Modify Zone/Subnet.

  4. In the Modify Zone/Subnet dialog box, select the checkbox and a vSwitch for an availability zone to add it, or clear the checkbox to remove it. Then, click OK.

    • When you select the checkbox for a target availability zone for a public-facing NLB instance, you must also associate an EIP with that availability zone. Select a newly purchased pay-as-you-go (pay-by-data-transfer) BGP (Multi-ISP) EIP with default security protection.

Change availability zone status

Availability zone statuses

Status

Description

Actions

Enabled

The NLB instance is enabled in this availability zone, and its virtual IP address (VIP) forwards traffic normally.

Remove DNS record

Note

If an NLB instance has only one availability zone in the Enabled state, you cannot remove its DNS record or stop the instance.

DNS Removed

The VIP of the NLB instance is removed from the NLB domain name resolution in this availability zone.

Restore DNS record

Remove and restore DNS records

You can remove and restore DNS records for the availability zones of an NLB instance in the NLB console. Use these operations to change an availability zone's status and simulate scenarios such as zone-level disaster recovery.

Limitations

Effective time

Billing impact

Use cases

You cannot remove the DNS record if the current NLB instance has only one availability zone in the Enabled state.

Removing or restoring a DNS record for an availability zone usually takes effect immediately. However, you may experience a delay of a few minutes due to network latency.

None.

Important

After you remove a DNS record, traffic to the VIP of the availability zone is still forwarded and incurs LCU fees.

Simulate scenarios such as multi-zone disaster recovery to verify the high availability of NLB.

  1. Log on to the NLB console.

  2. In the top navigation bar, select the region where the NLB instance is deployed.

  3. On the Instances page, click the ID of the target NLB instance.

  4. On the Instance Details tab, in the Zone section, perform one of the following operations.

    • Remove a DNS record

      In the Actions column of the target availability zone, click Remove DNS. In the dialog box that appears, confirm the impact and then click Remove DNS.

      Note
      • This operation affects only the current NLB instance. Other instances in the availability zone are not affected.

      • After the DNS record for the availability zone is removed, the A record that maps its VIP is deleted from the NLB domain name resolution. Evaluate the business impact before you perform this action.

      After the DNS record is removed, the status of the availability zone changes to DNS Removed, and the health check on the VIP of the availability zone stops.

    • Restore a DNS record

      To restore an availability zone whose status is DNS Removed, click Resume DNS in the Actions column. In the dialog box that appears, confirm the impact and then click Resume DNS.

      Note

      After the DNS record is restored, an A record for the availability zone's VIP is added to the NLB domain name resolution. The VIP then forwards traffic sent to the NLB domain name.

      After the DNS record is restored, the status of the availability zone changes to Enable, and the health check on the VIP of the availability zone starts.

Increase maximum public bandwidth

How maximum bandwidth is determined

A public-facing NLB instance uses its associated EIPs to provide services over the Internet.

The maximum public bandwidth depends on whether the NLB instance is added to an Internet Shared Bandwidth:

  • The NLB instance is not added to an Internet Shared Bandwidth: The maximum bandwidth of the public-facing NLB instance depends on the maximum bandwidth of the EIPs assigned to each availability zone. The maximum public bandwidth of each availability zone equals the maximum bandwidth of its corresponding EIP.

    NLB supports only pay-as-you-go (pay-by-data-transfer) EIPs that are not added to an Internet Shared Bandwidth.

    Important

    The maximum bandwidth of a single pay-as-you-go (pay-by-data-transfer) EIP is a reference value and upper limit, not a guaranteed service level agreement (SLA). Consequently, the sum of the maximum bandwidth values of all EIPs—the theoretical maximum bandwidth of the NLB instance—is also not guaranteed by an SLA.

    For example, consider an NLB instance that is deployed in three availability zones and is assigned an EIP in each zone. Each EIP has a maximum bandwidth of 200 Mbit/s. If underlying bandwidth resources are insufficient, the actual forwarding bandwidth of a single EIP may not reach 200 Mbit/s, and the total forwarding bandwidth of the NLB instance may not reach 600 Mbit/s.

  • The NLB instance is added to an Internet Shared Bandwidth: When you add an NLB instance to an Internet Shared Bandwidth, all EIPs associated with the NLB instance are also added. The individual bandwidth limits of the EIPs no longer apply, and bandwidth is managed by the Internet Shared Bandwidth. The maximum public bandwidth of the NLB instance equals the maximum bandwidth of the Internet Shared Bandwidth.

    Important

    The maximum bandwidth of an Internet Shared Bandwidth is a reference value and upper limit, not a guaranteed SLA.

How to increase maximum bandwidth

  1. If the bandwidth of an EIP in an availability zone is less than the maximum value of 200 Mbit/s, you can upgrade the EIP to increase its bandwidth to the maximum value.

  2. If your requirements are still not met, you can use the following methods to further increase the public bandwidth:

    1. Add availability zones to the NLB instance: In regions that support multi-zone deployment for NLB, you can add more availability zones to the NLB instance to assign more EIPs and increase the overall maximum public bandwidth. This method is ideal when you want to minimize public network costs.

    2. Add the NLB instance to an Internet Shared Bandwidth: An NLB instance supports only one EIP per availability zone. Therefore, the theoretical maximum bandwidth of a single NLB instance = maximum number of NLB availability zones × 200 Mbit/s (theoretical maximum bandwidth per EIP). If your requirements exceed this theoretical limit and you can accept higher costs, you can add the NLB instance to an Internet Shared Bandwidth in the console.

Potential causes of packet loss

  • Inconsistent EIP bandwidth across availability zones: A public-facing NLB instance uses DNS round-robin to distribute traffic evenly among its EIPs for zone-level disaster recovery. If EIPs in different availability zones have different bandwidths, traffic shifting from a high-bandwidth EIP to a low-bandwidth EIP can cause packet loss.

    For example, you have an NLB instance deployed in three availability zones, and the EIPs are configured with maximum bandwidths of 200 Mbit/s, 200 Mbit/s, and 100 Mbit/s, respectively. Assume all EIPs can reach their theoretical maximum bandwidth. When the total traffic exceeds 300 Mbit/s, each EIP is allocated more than 100 Mbit/s of traffic. In this case, traffic distributed to the 100 Mbit/s EIP will experience packet loss.

  • Uneven traffic distribution: Even if the EIPs in multiple availability zones have the same maximum bandwidth, packet loss may occur if traffic is unevenly distributed among them. For example, this happens when clients access the NLB instance by using an EIP directly.

    For example, consider an NLB instance deployed in three availability zones, and the EIPs are configured with maximum bandwidths of 200 Mbit/s, 200 Mbit/s, and 200 Mbit/s, respectively. Assume all EIPs can reach their theoretical maximum bandwidth. Due to uneven traffic distribution caused by clients accessing an EIP directly, the traffic to the three EIPs is 200 Mbit/s, 100 Mbit/s, and 0 Mbit/s, respectively. At this point, any further traffic sent by clients to the first EIP will experience packet loss.

  • Insufficient backend server capacity: Even with sufficient bandwidth, if backend servers cannot handle high-concurrency requests in a timely manner (for example, due to CPU overload, response delays, or TCP retransmissions), indirect packet loss can occur.

Configuration recommendations

  • When configuring the maximum bandwidth for EIPs or an Internet Shared Bandwidth, set it slightly higher than your business requirements.

  • When you use EIP bandwidth:

    • Assign EIPs with the same bandwidth to each availability zone of the NLB instance to prevent service disruptions caused by bandwidth differences during a failover.

    • Access the service by using the NLB domain name instead of an EIP directly to avoid uneven traffic distribution.

References