Secret Server Cloud: US - RDP/SSH Connectivity Issues

Incident Overview

Customers in the US region experienced intermittent failures when initiating RDP and SSH sessions through Secret Server Cloud, specifically via the Connection Manager / RDP proxy.

Start Time: January 23, 2026, 5:45 PM ET
End Time: January 23, 2026, 6:45 PM ET

Some customers observed:

• Failures when launching RDP or SSH sessions
• Black screens during connection attempts
• Inability to connect via the RDP proxy, while direct RDP access may have continued to work

Other regions were not impacted.

Detection

The issue was initially identified through customer support tickets. Internal alerts were received later due to alert thresholds being set too high to detect the early stages of the degradation.

Root Cause

As part of a planned maintenance activity in the US region on January 23, customer-facing Azure Service Bus resources were migrated to new hostnames, along with updates to email services and domains. The maintenance completed at 4:30 PM ET.

To avoid data loss and disruption to active customer sessions during the maintenance:

• A replica set of compute nodes continued to point to the legacy Service Bus
• A new set of compute nodes were provisioned to use the new Service Bus hostnames

This temporary configuration required approximately double the normal compute capacity, which is not a typical operating condition. As a result:

• Some compute nodes remained in a provisioning state longer than expected due to resource constraints in the node pool
• Over the following hour, insufficient available capacity existed to process incoming connection requests
• This led to RDP and SSH session failures and disconnections via the Connection Manager

Once the issue was identified, capacity was increased by:

• Expanding the compute node pool
• Rebalancing resources away from the replica nodes pointing to the legacy Service Bus

These actions allowed the backlog of requests to be processed and restored normal service behavior. This behavior was not observed in lower silos or in other regions during earlier phases of the maintenance period.

Preventive Actions

To prevent recurrence and improve detection during atypical load conditions, the following actions are being implemented:

• Lower alert thresholds for message queue depth and processing delays to enable faster detection and response
• Reduce alert lookback periods to surface emerging issues sooner during rapid load changes
• Introduce a new monitor tracking the percentage of compute pods in an unhealthy state due to resource constraints in the node pool