8.2.30 Release Notes

Fixed an issue where IP rules for a Site-to-Cloud mapped connection could be erased when the gateway started up.

Fixed an issue where disabling Distributed Cloud Firewall (DCF) on a transit gateway with Site-to-Cloud (S2C) or other external connections could leave existing DCF filtering rules in effect on the interface when no policy change was queued at the time, allowing traffic to continue to be filtered by DCF after DCF was disabled.

Fixed an issue where new gateway deployment or software/image upgrade for an existing gateway could fail if a container responsible for gateway-to-Controller secure communication did not start within the default 15-minute initialization window.

Fixed an issue where NAT rules could fail to install on gateways with legacy tunnel peerings that predated recent tunnel-peering enhancements.

Fixed an issue where Automatic Controller Access Security Group Management enforcement could unintentionally disable Security Group Management when a transient error occurred during an enforcement check. When Security Group Management became disabled, all gateways experienced keep-alive failures and new gateways could not come up.

Fixed an issue where a Controller upgrade from 8.1.x to 8.2.20 or 8.2.30 could fail during data migration with an IP-pool allocation error, aborting the upgrade.

Fixed an issue where Azure Security Group Management could unexpectedly delete and recreate a security group rule during gateway launch when an Azure API throttling response on a rule-existence check was misinterpreted as the rule not existing.

Fixed an issue where the Controller’s GCP event monitoring could fail to query event logs for a GCP account when that account’s associated service account was linked to a very large number of GCP projects, causing the query to fail with an error indicating the event filter was too long.

Fixed a rare deadlock in the internal policy-update path on gateways enforcing Distributed Cloud Firewall (DCF) policies that include SmartGroups with FQDN selectors. The deadlock could previously cause the gateway data-plane process to stop forwarding traffic while remaining alive. The gateway process did not exit and was not marked unhealthy by existing watchdogs. This was more likely when the FQDN selector resolved to a set of addresses that rotated frequently (short DNS TTL), because each resolution change triggered a policy reload that raced with an internal hit-counter update. Recovery previously required manually restarting the affected gateway. WebGroup (SNI-based) policies and gateways not using FQDN-selector SmartGroups were not affected by the deadlock.

Creating a new gateway with the same name as an existing gateway may cause local files of the existing gateway to be deleted when the creation fails. The existing gateway name disappears from the Controller CLI once this occurs. This can break SSH access to the existing gateway.

Impact:

Affected Scenario:

Gateway creation using a name that already exists.

Workaround:

Do not reuse gateway names. Contact Aviatrix Support if recovery is required.

When using the update_spoke_vpc_route_table API to onboard an Azure route table, the default route (0.0.0.0/0) is not programmed in the spoke VNET route table if the Spoke Gateway has learned the default route from any of the following sources:

In all of the above cases, although the Spoke Gateway has successfully learned and installed the default route in its own routing table, the route is not re-programmed into the associated Azure VNET route table during the onboarding operation.

Impact:

Workaround:

Manually add the default route to the Azure route table. Contact Aviatrix Support for assistance.

In Azure environments, when Distributed Cloud Firewall (DCF) Security Group Orchestration attaches a Network Security Group (NSG) to a subnet, the subnet name is changed to all lowercase. Although Azure resource names are generally case-insensitive, this modification causes issues with Infrastructure as Code tools such as Terraform, which treat resource names as case-sensitive. Terraform may flag affected subnets for replacement, potentially disrupting existing deployments.

Impact:

Workaround:

In Terraform, add a lifecycle block with ignore_changes for subnet_id (as well as id in the azurerm_subnet resource) to prevent forced resource replacement. Note that this workaround does not restore the original subnet name casing.

Gateway launch may fail with a tls: bad certificate error when pulling container images. The Controller’s registry TTL eviction (garbage collection) may fire while images are being downloaded to the gateway, corrupting in-flight blob transfers. The gateway’s container initialization cannot complete, and an internal identity-helper service on the gateway exits with code 125.

Impact:

Workaround:

Restart the Controller’s core application service to reset the registry state, then retry gateway creation. Contact Aviatrix Support for assistance.

Some gateways may become stuck in a failed configuration-push state for an extended period after an Image Upgrade to Aviatrix 10.0.10 (from 9.0.10). This has been observed primarily on lower-memory gateway instance types (for example, AWS t3.micro).

Impact:

Affected gateways may show a "Not Up-To-Date" config status and a down backup tunnel hours after the image upgrade completes. The root cause is a Distributed Cloud Firewall (DCF) configuration push failure caused by insufficient available memory on memory-constrained instance types.

Workaround:

Retry the gateway Image Upgrade. If the gateway remains stuck in this state, increase the gateway’s instance size for additional memory headroom, or contact Aviatrix Support.

On Controller versions prior to 10.1.0, it is possible to onboard multiple Aviatrix access accounts that map to the same underlying cloud service provider (CSP) account. Because several Controller features assume a one-to-one relationship between an Aviatrix access account and a CSP account, duplicate access accounts can lead to unpredictable configuration and operational behavior. Duplicate is defined per CSP as follows:

Starting in Controller 10.1.0, the Controller blocks new onboarding of duplicate access accounts across all supported CSPs. Existing duplicate access accounts continue to operate as-is and are not automatically removed.

Impact:

Workaround:

Onboard only one Aviatrix access account per underlying CSP account. If duplicate access accounts have already been onboarded and are causing issues, contact Aviatrix Support for assistance with identifying and removing the duplicates.

A VPN user’s profile assignment could appear as N/A immediately after the user was attached to a profile through the API, until the user’s next reconnection.

Impact:

Workaround:

Have the affected user reconnect to update the displayed profile assignment. Contact Aviatrix Support for assistance.

Enabling both Policy-Based Routing (PBR) and NAT logging under User VPN settings can cause high CPU utilization on the User VPN gateway during data-plane traffic. The high CPU usage occurs because NAT log entries generated by PBR are written to the gateway’s serial console.

Impact:

User VPN gateways with both PBR and NAT logging enabled may experience elevated CPU utilization while passing data-plane traffic, which can degrade VPN gateway performance.

Workaround:

Disable NAT logging on the affected User VPN gateway. If NAT logging must remain enabled, contact Aviatrix Support to apply a backend workaround.

During a Gateway resize operation (for example, changing the Gateway’s underlying instance or VM size), the Gateway can experience a temporary period of traffic loss instead of a seamless transition. This has been observed during testing on GCP, AWS, and Azure Gateways, where a resize between certain instance sizes produced approximately 2 minutes of packet loss.

Impact:

Workaround:

No workaround is currently available. A fix is planned for a future 10.1.x maintenance release.