Parallels released Parallels RAS Core v21.2.1-27306 on 30 July 2026. Billed as a maintenance release, it actually has a very distinct character: out of nine changelog entries, three cover Hyper-V and three cover printing. The rest touches identity (Enrollment Server, Web Authentication) and load balancing.
Here is what it changes in the field, and who should patch first.
The new feature: printing preferences exposed in the User Portal
Document printing preferences set inside a remote session — including printer tray selection — are now exposed when printing through the RAS Universal Printer in User Portal sessions, for third-party clients and extensions.
In practice: until now, a User Portal user selecting a specific tray from a published application would see that choice ignored or passed inconsistently to the local printer. This is the classic scenario in environments with letterhead in tray 1 and plain paper in tray 2 — accounting firms, law firms, HR departments. The selection now makes it through.
Worth noting: this exposure explicitly targets third-party clients and extensions, meaning integrations built on top of the User Portal rather than the native Parallels Client.
The critical fix: Provider Agent crash
This is the only item flagged as a Critical Fix in the build. The Provider Agent could crash in three situations: during Hyper-V management, during VDI cloning, or while clearing active guest sessions as a host shuts down. The reported causes are intermittent access errors and connection resets.
Why it matters: the Provider Agent is the component that talks to the hypervisor. When it goes down, the broker loses visibility over the VDI pool — no provisioning, no power management, and hosts left in inconsistent states. On infrastructure that clones during business hours to absorb load, that is an immediate functional outage.
The word intermittent matters here: this class of crash does not reproduce on demand, it shows up under load or during maintenance windows. If you have unexplained restarts of the RAS Provider service in your event logs, you probably just found your explanation.
Hyper-V Cluster: two fixes that save you manual cleanup
Two separate corrections, telling the same story — Hyper-V cluster handling had room for improvement.
Host deletion leaving orphaned resources
Deleting Hyper-V Cluster hosts could fail sporadically, leaving orphaned virtual machines or disk resources behind in the hypervisor. This is the kind of bug that slips past unnoticed: the RAS console shows the host as removed, the administrator moves on, and the VHDX files keep occupying the CSV for months. On a VDI pool that recreates regularly, the build-up is fast.
If you are running an earlier version, it is worth auditing your CSVs before patching: compare the machines the broker knows about against what the cluster actually holds.
Cloning onto unavailable nodes
VDI host cloning in a Hyper-V Cluster could fail when the system attempted to place new hosts on nodes that were paused, isolated, or quarantined.
This is a textbook maintenance scenario: you pause a node to apply patches, RAS keeps treating it as a valid placement target, and provisioning breaks. The fix is to account for the node’s actual state. In practical terms, draining a node for maintenance no longer needs to be coordinated with suspending provisioning on the RAS side.
Universal Printing: correct trays and AVD support restored
Tray names that finally mean something
Printing preferences could display generic or irrelevant tray names when a redirected printer had no specific client-side preferences. From the user’s point of view: labels matching nothing physical, and a support ticket every single time.
The Paper Source, Paper Form and Print Quality options now correctly match the local printer settings.
Universal printing on AVD
More drastic: RAS Universal Printing simply failed on Azure Virtual Desktop resources. For hybrid infrastructures publishing both on-premises RDSH and AVD, that meant a two-tier printing experience depending on where the resource came from. Now fixed.
Identity: Enrollment Server and Web Authentication
Enrollment Server: silent directory pollution
The nastiest fix in the build. If the Certificate Authority is unavailable when the Enrollment Server starts, the server could generate an excessive number of certificates in Active Directory, and fail to clean up the associated private keys.
That is two problems in one. First, certificate object proliferation in the directory, bloating user objects. Second, private keys left behind on the server with no usable certificate attached — accumulation that has no business being there, on a component that is sensitive by definition.
The trigger scenario is mundane: an Enrollment Server restart while the CA is down — maintenance, a grouped reboot after Patch Tuesday, a network outage between the two. If you run smart card logon or Web Authentication, this fix alone justifies the upgrade.
Web Authentication: the error that isn’t one
Web Authentication could display an incorrect username or password error with newly issued certificates, even though login succeeded once the prompt was dismissed.
Technically minor, operationally irritating: this is exactly the kind of behaviour that generates support calls and erodes user trust in the platform. A user trained to ignore one error will ignore the next one too — including the one that actually matters.
Load balancing: multiple assignment when the buffer runs dry
Users could be incorrectly assigned multiple hosts within the same host pool when launching an application, if no hosts were available in the buffer.
The scenario: the buffer is empty — morning connection spike, provisioning lagging, or an undersized pool — and the assignment mechanism hands several machines to the same user instead of one. The result is hosts consumed for nothing, which worsens the very shortage that caused the problem. A classic snowball.
If you size your buffers tightly to keep costs under control — typically on Azure — this one counts.
Full release notes are available in KB 131037.
