Added the ability to backup/restore the configuration, and/or migrate the entire R2FS! node to a new machine/VM

This is the last of 9-step process that completes the transformation of our HA (high-availability) subsystem, which we begun a few months ago; this step added 25 massive scale tests to our CI/CD pipeline, covering edge cases like millions of users/VFSs and hundreds of virtual sites on dozens of cluster nodes

The above test suite allowed us to find and fix one latent bug that only affected large scale HA deployments (the fix is confirmed live, in production, by large customers who were given early access to v8.1.4)

Fixed a UI bug which prevented the R2FS! node identity key from being displayed if you created/edited an admin account before creating your first VFS configuration

Fixed a bug in the worker process that prevented proper propagation of “shared object links” to the other nodes of an HA cluster

Usual small set of little bug-fixes and quality of life improvements

This is a PERFORMANCE RELEASE, designed to take full advantage of the stream multiplexing and buffer auto-tuning in R2FS! v2.2.1.
If you use an R2FS! backend and are used to “capping it out” at around ~1 Gbps, this version unleashes transfer speeds up to 90x better, provided your network and back-end storage can keep up with it 😉

Usual small set of little bug-fixes and quality of life improvements

We call this our PERFORMANCE RELEASE. We added stream multiplexing and auto-tuning of memory buffers (configurable individually per-VFS). The results are mind-blowing: we went from a maximum throughput of ~1.08 Gbps (no matter how fast your network and storage were) to saturating a 10 Gbps network without even pushing it on real hardware here at our offices (and our estimates show that ~88 Gbps over a 100 Gbps network should be easily attainable).
NOTE: requires Syncplify Server! v8.1.2 or newer.

Added 6 new bounds-checkers and 2 more “panic recovery” paths ◀ this is the key improvement in v8.1.1 which make this version far more stable than any version before it

Hardened a few code paths to better deal with Windows-specific inability to include stderr output in its Event Log (our Linux version was never affected by this bug because it’s OS-specific)

Fixed a problem with VFS quota-updater that could occasionally blank-out remote/cloud VFS credentials upon saving the VFS configuration

Fixed a WebClient! UI problem with the shared-object password field relying on the browser’s credential storage for auto-fill

Usual small set of little bug-fixes and quality of life improvements

Added an optional checkbox to the Admin UI to enable mandatory strict key-pinned TLS negotiation with R2FS! v2.2+; enabling this option will require R2FS! v2.2 or newer (just released today) and fail when the R2FS! version you’re running is below that release number

Fixed the shared-object link sent in notification emails which mistakenly did not contain the protocol, host and port

Fixed semantics of “same IP on PASV” which was interpreted backwards

Added an auto-recoverer and improved channel-open correctness tests in the SSH (SFTP) listener

Dropped reliance on the system’s shell (sh, bash, cmd, powershell, …) and implemented a fully self-contained (and jail-rooted) minimal shell focused entirely and exclusively on a small set of file-related commands (after all this is a file-transfer server, it was never meant to be a replacement of OpenSSH for complete remote system management); in this v-shell type help to see a list of available commands

Usual small set of little bug-fixes and quality of life improvements

Added a stricter pinned-key TLS mode, designed to work in conjunction with Syncplify Server! v8.1+ (which is itself scheduled to be released later today); when enabled R2FS! will refuse to connect to Server! versions prior to v8.1 to prevent the possible adoption of a more relaxed TLS enforcement
NOTE: even with this setting disables, when R2FS! v2.2+ connects to a Server! v8.1+ the stricter TLS mode is always guaranteed to be enforced

Identity verification during HA-JOIN is now mutual and bidirectional

Increased password hashing rounds to the levels required for full FIPS 140-3 compliance (and all existing hashes are automatically upgraded at next login with zero administrative burden)

Fixed a bug that allowed a SuperAdmin to create a password-less SuperAdmin account (the password-less SA would be dead-on-arrival though, as password-less SuperAdmin logins are always forbidden by the API, so this does not constitute a widening of the software’s security posture)

Strengthened TOTP replay prevention in all 2FA/MFA paths, now even cross-node when the software runs in high-availability (HA) deployments

Improved the TOTP general user experience (UX), now featuring recovery codes (shown only once during enrollment)

Added the ability for each individual user to revoke “trusted devices” (specifically for the TOTP authentication path)

WebClient! now properly honors the “force change password” and “force enroll in MFA” settings when an Admin sets them for a specific user account

Fixed a bug that only affected R2FS!-type Virtual File Systems (VFSs) and prevented the Content-Disposition header from being properly formatted upon downloading files from the WebClient! UI. Special thanks to our friend Jeff, not only for finding and reporting this, but also for helping us reproduce the issue in our own systems, which ultimately made the fix nearly trivial.

Fixed a UI regression just introduced in v8.0.7 yesterday, which essentially capped the number of users viewable in the Admin UI to 1,000 (this hot-fix restores the full uncapped user list)

Fixed several issues with the new web UI that was introduced with v2.1.0, including uploading certificates, CORS, security headers, and more

Fixed a backend glitch that would equate Any operation to No operation in the OpLog

Updated the Go compiler and standard library to the most recent version, which comes with its own internal small bug fixes

Fixed a glitch in the web UI which prevented the creation of new admin profiles erroneously reporting a “missing password hash” (it was a UI-only issue, not a password issue)

Fixed a bug in the logging subsystem when outputting colored logs to a non standard tty

Updated the Go compiler and standard library to the most recent version, which comes with its own internal small bug fixes

This is the 4th and last small step that completes our high-availability (HA) total redesign; functionally it’s identical to all other V8.0.x that preceded it, but the SuperAdmin UI now has gained the ability to automatically suggest correct bindings for each Node of each Virtual Site upon joining an HA cluster

Usual small set of little bug-fixes and quality of life improvements

Fixed an edge-case bug in SyncJS’ RunAsync() function, which on POSIX systems could leave behind a zombie process if used to invoke a shell with a command line to execute a shell-script (very uncommon, but possible nonetheless)

Added CtxVirtualSite() to the SyncJS scripting engine

Usual small set of little bug-fixes and quality of life improvements

Added a two-fold protection shield against Windows’ IRP leakage bug when remote storage accessed via SMB/CIFS becomes unavailable mid-session (this is a Windows bug, not ours, but we can now work around it)

Added CtxUsername() to the SyncJS scripting engine

Usual small set of little bug-fixes and quality of life improvements

Fixed a bug in our Google Cloud Storage VFS implementation that prevented certain file-transfer clients from correctly unmarshalling directory names

Added CtxVFSName() and EventCtx() to the SyncJS scripting engine

Usual small set of little bug-fixes and quality of life improvements

Added support for Passkey sign-in to all UIs (SuperAdmin, Admin, and of course WebClient!) to include yet another strong authentication mechanism to the already rich set supported by Syncplify Server!

Page-wide drop-zone to upload files and folders with a more modern user experience similar to Google Drive, OneDrive, etc… just drop the items you want to upload anywhere and they’ll be uploaded into the current directory displayed in your WebClient! UI at that time

Fixed a compatibility issue in our OIDC implementation that prevented login completion when connected to Microsoft Entra ID (note: all other OIDC providers were already working well, this issue was exclusively related to the way the optional “alg” claim is handled, which Entra ID does differently from any other OpenID Connect IdP)

Usual small set of little bug-fixes and quality of life improvements

Improved accessibility (WCAG 2.2 AA) and usability (UX) of the WebClient! UI with a small set of targeted optimizations

Added an all-new CtxRelTargetPath() to our scripting engine, specifically designed to retrieve the destination (target) relative-path inside of the AfterFileRename event-handler

Backups taken from nodes that used to be part of an HA cluster and then they left and went back to being stand-alone single servers now no longer contain legacy HA node bindings

Usual small set of little bug-fixes and quality of life improvements