26.10.0 Quarterly Release

Changes compared to 26.7.3
Notice
- Microsoft 365: OneDrive now follows your selection on full backups. Before 26.10.0, full backups could capture a user's personal OneDrive even when the OneDrive service wasn't selected. Now both full and incremental backups follow the selection, so unselected OneDrive data won't be in new backups, and the job log shows a warning when content is skipped. Check your Microsoft Office 365 Protected Items. See Office 365 - Configuring Selections.
Behavior changes to know about
- File and Disk Image restores no longer recreate the original folder path under your destination. Selected items land directly in the folder you pick. Update any restore runbooks or scripts that expect the full path
- Microsoft 365 restores skip items that already exist at the destination. This is on by default
- Policy backup rules for SMB, WebDAV or Storj Storage Vaults now create schedules. If you have policies like this, new scheduled jobs will start running after the upgrade
- Microsoft 365 mailbox restores now use the Microsoft Graph API instead of Exchange Web Services (EWS)
- The backup app now runs Storage Vault search, email preview, retention passes, index rebuilds, and snapshot and file deletion through the background service, instead of as the signed-in user
- The backup app now asks the user for consent when a policy change requires it
- Storage Vault size stats measured during retention passes now use exact on-disk sizes, so reported sizes may shift slightly after the upgrade
- Saving an organization's settings restarts its webhook delivery, and queued deliveries are sent again from the start, so you may see repeats
- Registering a device with a device ID that's already registered is now rejected with "This device is already registered" instead of overwriting the existing device
- Admin sign-in with a security key now requests user verification (PIN or biometrics) as "preferred"
New Features
Restores
- Restore a Protected Item, or a whole snapshot, in your browser from the Web Portal as a download, with no app needed. A new Allow administrator to restore data setting lets a client let you do it for them
- Restores can write straight to remote and cloud storage locations, not just a local disk
- Mount a Files and Folders snapshot as a read-only folder on Linux, including in WSL2
- Skip mail, OneDrive and SharePoint items that already exist when restoring to Microsoft 365 (on by default)
- A Test Connection button for Proxmox restores in the Web Portal, plus SSH host key options
Integrations
- NinjaOne RMM: backup status in device custom fields, plus optional tickets that open, escalate and close automatically
- ConnectWise Manage PSA: tickets for failed and missed backups, matched to the client, device and Protected Item
Backups
- Microsoft 365 account selection now uses include and exclude rules, with an All accounts rule for new users (devices need 26.8.1 or later)
- Speed limits can follow a schedule, for example slower during business hours and full speed overnight
- NFSv4 is supported for Storage Vaults
- User profiles have an Extra Fields section for notes, contact details or IDs, each one admin-only or visible to the user
Administration
- Admin API tokens let you use the API without sharing your password or 2FA
- A new Restrict this account to read-only access admin option blocks an admin account from any API action that makes changes
- A Show advanced actions option in the Device Actions dialog
- MySQL Protected Items can save the server's user accounts and privilege grants as a SQL script, and store an integration's recovery metadata file in each snapshot. Database replay restores skip both; get them back with a file restore
Recovery media
- Build the Linux recovery ISO on Windows with WSL2, with no Docker or Podman needed
- The Windows recovery environment has its own Disk Management tool for preparing disks before a Disk Image restore
Enhancements
- Faster Storage Vault index loading that uses less memory, especially for large vaults
- More reliable Linux Disk Image backups: LVM snapshots grow automatically, snapshot data stays on disk instead of in RAM, and failures are clearly explained in the job log
- Safer Disk Image restores: you're asked to confirm before restoring onto a mounted volume, and the running OS disk can't be overwritten
- Faster, lighter Microsoft 365 backups and mail restores. A SharePoint or OneDrive site that fails an incremental backup now falls back to a full backup instead of being dropped
- Much less RAM needed to browse or restore a single snapshot
- Restored items now land directly in the folder you choose
Other improvements
- Less load on the Web Portal when jobs are scheduled frequently
- Better cancellation of backup and restore jobs on FTP Storage Vaults
- Seamless redirection when signing in to the Web Portal with external OAuth credentials
- Microsoft 365 mailbox and SharePoint/OneDrive backups share one worker pool, with a guaranteed share for each
- Faster backup jobs while a retention pass is running, and faster restore startup when only a few files are selected
- Clearer job log errors when a VMware vSphere restore fails creating or registering a VM, and when a Hyper-V backup fails creating a checkpoint
- Duplicating a Protected Item in the app now selects the new copy
- Unlocking a Storage Vault removes damaged lock files, and reports an error instead of success if some locks couldn't be read
- Updated the OpenSSL library used for VMware support on Windows
- Clearer admin account permissions in the Web Portal, and safeguards against two people changing a user profile at the same time
- The Proxmox restore Test Connection in the app prompts for a destination node and storage target if they aren't selected yet
- Updated Portuguese (Brazil) translation
Bug Fixes
Microsoft 365
- Fixed incremental backups of SharePoint sites or Teams leaving out document libraries when SharePoint was selected but OneDrive wasn't. Only the first backup was complete. Every later backup missed the libraries while still reporting success. Affected items fix themselves on the next backup, but snapshots taken while affected don't contain those libraries
- Fixed duplicated calendar events, and mail folders landing in the mailbox root, on Microsoft 365 restores
- Fixed a simulated Microsoft 365 restore reporting that no files were restored
- Fixed automatic setup of a China-region tenant saving the wrong region, which made later backups fail
Restores
- File and Disk Image restores no longer recreate the full original folder path inside your destination. Selected items land directly in the folder you choose, and the destination folder's permissions and timestamps are left alone
- A Disk Image restore to physical disks could overwrite the disk the running operating system is on, crashing the device partway through the restore. It now stops before starting and tells you to boot from recovery media, and the system disk is greyed out in the Restore dialog
- Fixed granular restores hanging on Magnus Box Cloud Storage
- Fixed missing options when running a granular restore in simulate mode
- You now get a clear error when you start a restore remotely on a device whose backup agent is too old for that restore type
- Fixed the Restore wizard in the app jumping to the wrong page
- Fixed restores to VMware vSphere through vCenter failing with "do not belong to the same compute resource" when the datacenter has more than one cluster or standalone host
Proxmox
- Fixed many issues editing Proxmox Protected Items in the Web Portal: disks numbered 10 and above were skipped, private keys, host key settings, exclusions and connection details could be lost on save (especially with the device offline), and the Password/Private key fields didn't match the saved mode. The Private key box is now multi-line
- Proxmox SSH host key checks: before this release, editing a Proxmox Protected Item in the Web Portal reset host key verification to Allow any host key on save, and Proxmox restores started from the Web Portal never checked host keys. The Web Portal now has host key verification options for both. Action: re-check the host key setting on your Proxmox Protected Items after the upgrade
- Fixed the app crashing when opening a Proxmox Protected Item after one of its disks was detached from the VM
- Proxmox restores now check that a server address, node and storage target are set before starting, instead of silently never starting or failing partway through
- Fixed Proxmox to Hyper-V/VMware restores failing with "mbr bad signature" (disk with no partition table) or when a CD-ROM ISO was attached
- Fixed Proxmox backups where, if merging the temporary copy-on-write overlay back into the base image failed, the overlay stayed attached to the VM and the failure wasn't reported. The merge is now retried, and a failure is reported
- Fixed the "All VMs" label in the Web Portal Proxmox selector not being translated
Hyper-V and VMware
- Fixed Hyper-V, Proxmox and VMware backups losing incremental speed after "not present in any index" warnings
- Fixed VMware vSphere backups on Windows ignoring a custom port in the Server field when downloading virtual disks
- Fixed the guest quota warning not showing when you first open a VMware or Hyper-V item set to back up all VMs
- Fixed the connection test for VMware, Proxmox and MongoDB items saying it worked when the server couldn't actually be reached
Backup jobs and Storage Vaults
- Fixed jobs not retrying in some cases when other work was scheduled around the same time, and duplicate-file errors during backup are now handled cleanly
- Fixed an issue with detecting write failures to a Storage Vault
- Fixed "Skip if already running" occasionally starting a duplicate job, extra "Failed to check for running jobs locally" errors, and unreadable lock files being ignored when checking if a vault was in use
- Fixed a finished backup job sometimes showing as running again
- Fixed the automatic out-of-memory retry (using temporary files instead of RAM) not running for SQL Server, Exchange, Hyper-V, Windows Disk Image and VMware jobs, and retention passes being wrongly retried as out-of-memory failures
- Fixed jobs set to "Prefer temporary files instead of RAM (slower)" failing to find existing data when retention ran at the same time
- Fixed handling of network or server outages on FTP, WebDAV, Storj and spanned Storage Vaults, and uploads of large files to WebDAV
- IDrive e2 vaults no longer look up their region on every connection, so jobs don't hang for up to 20 minutes when that lookup is slow
- Fixed policy backup rules for SMB, WebDAV or Storj vaults not creating a schedule. Existing policies will now create them
- Fixed a connection leak in the backup agent when uploading a job log kept failing: each retry left a connection open
Backup app and Web Portal
- Fixed macOS devices losing Full Disk Access after automatic updates
- The app now shows an error when the background service refuses to start a job, instead of showing nothing
- Fixed "CreateFile failed: 231" errors in the app
- Fixed two Protected Items with the same name swapping places in the sidebar, and snapshot dates showing in the wrong time zone
- Fixed odd labels on some browse buttons when the app runs in a language other than English
- Client installation errors now show when you're signed in to the Web Portal as an admin
- Fixed admin sign-in with newer FIDO2 security keys that need a PIN or biometrics
- Fixed the Web Portal pausing for 10 seconds when opened from another site, and invisible button text in light mode
- Fixed inconsistent tree views in the Web Portal
- Fixed searching for "Email" in the Web Portal opening a blank page
- Fixed custom branding making active tab titles and sorted column headers hard to read in dark mode in the Web Portal
API, webhooks, email and AI connection
- Fixed AI assistant (MCP) tools not returning job logs correctly in some cases, and improved the instructions our MCP server gives to AI assistants so they use the tools more accurately
- Webhook delivery workers are now stopped properly when webhook targets change. Saving an organization's settings now restarts its webhook delivery, so any queued deliveries for that organization are sent again from the start, even if the targets didn't change
- Registering a device with a device ID that's already registered no longer silently overwrites the existing device. The request is now rejected with "This device is already registered"
- Security fix: changing a user's profile or resetting their password from a device or end-user session no longer returns admin-only profile notes or other devices' Protected Item settings in the response
- Admin API requests for an organization or user group you don't have access to now return a proper "not authorized" (403) error instead of a server error
- Security fix: added protection against server-side request forgery (SSRF) when email is sent through a custom SMTP server
Plus 20 server-side fixes that we apply on our servers, including more reliable automatic Storage Vault unlocking and key rotation. You don't need to do anything for these.