Underlying Ceph Upgraded to the 19.2.6 Security Release

On August 19, 2026, the Ceph project released Squid 19.2.6, a security release that fixes four CVEs. The Ceph project strongly recommends that all operators upgrade as soon as possible. UniVirStor 4.19.7 ships with Ceph 19.2.6, so both newly deployed clusters and upgraded existing clusters run this version.

The Four CVEs Fixed in This Release

CVSS scores are taken from the official Ceph security advisory; other vulnerability databases may rate these CVEs slightly differently.

CephX Key Rotation: What 4.19.7 Covers

CVE-2025-30156 cannot be fixed by upgrading alone. Ceph introduces a new key type, aes256k, for this fix (the first time CephX has added a key type), and existing keys must be rotated to the new type to complete the remediation. The 4.19.7 upgrade procedure includes key rotation, with the following scope:

Why client.admin Temporarily Keeps the Older Key Type

Clients that mount CephFS or RBD through the Linux kernel (kernel clients) need Linux kernel 7.0 or later to support aes256k, and the kernels shipped by most distributions used in enterprise environments do not support it yet; the default Ubuntu 22.04 kernel is one example. Rotating client.admin to aes256k now, or disabling the older aes type, would immediately prevent these clients from mounting and interrupt service.

4.19.7 therefore migrates in stages: key rotation on the cluster service side is completed now, while rotating client.admin and disabling the older key type wait until client kernels have been upgraded. Until then, existing clients continue to work as before, with no need for an immediate kernel upgrade.

Full remediation of CVE-2025-30156 ultimately requires every client to support aes256k. We recommend starting to inventory the hosts that mount CephFS or RBD through the kernel and planning kernel upgrades; once that is done, Ambedded support is happy to work out a client key migration schedule with you.

How the Upgrade Works

4.19.7 is installed with an update package. Hosts are upgraded and their keys rotated one at a time, and the cluster health is confirmed after each host before moving on to the next. Ceph services restart in turn during the upgrade, so we recommend scheduling it within a maintenance window. Existing customers can contact Ambedded support to receive the update package and upgrade guidance.

Full CephX User Capability Management in the Web UI

Previously, CephX user capabilities in UVS Manager could only be set for OSDs. Setting MON, MGR, or MDS capabilities at the same time, or using Ceph's built-in profiles, required the command line. 4.19.7 introduces a redesigned capability editor.

Separate Settings for Four Service Types

When creating or editing a CephX user, you can set capabilities separately for mon, mgr, osd, and mds, and leave any type you do not need empty. A new user can also be created without any capabilities and have them added later.

Profiles and Advanced Parameters

In addition to allow r/w/x combinations, Ceph's built-in profiles are supported, along with parameters such as pool, namespace, and path. Common examples:

Existing Settings Fully Preserved When Editing

When you edit an existing user, the editor loads all of its current capabilities, including profiles and their pool, namespace, and path parameters, without truncating or dropping anything outside the OSD settings.

More Accurate Management and Monitoring

Faster CephFS Pages

The CephFS management page now uses a more efficient data source. Volumes, subvolume groups, and subvolumes share one set of data instead of querying repeatedly, so the page loads faster in environments with many subvolumes.

Disk Capacity in the Drive Position Panel

The disk panel in Server Hard Drive Position Settings now shows each disk's capacity alongside its serial number, making it easier to verify the drive layout. When the capacity cannot be read, the panel shows "-" rather than a misleading 0.

More Reliable CRUSH and iSCSI Settings

If an error occurs while renaming a CRUSH bucket type, the UI now correctly reports the failure instead of reporting success, and names are restricted to letters, digits, underscores, and hyphens. The iSCSI IQN Base Prefix is now validated and no longer accepts uppercase letters, so problems are caught before a target is created.

More Accurate Samba Alerts

The metric that tracks Samba service readiness could previously get stuck on an abnormal value when a system component stopped responding, and would not update after the service recovered. 4.19.7 adds timeouts to these checks, so the metric returns to normal at the next check after recovery. The SambaExporterDown alert is fixed as well: clusters without Samba no longer receive constant false alarms, and the alert now fires correctly when the Samba container stops.

More Efficient Alert Notification Handling

When an alert arrives, the system no longer makes an extra Prometheus query to look up rule details, and when email notification is disabled, the rest of the processing is skipped. This reduces system load when many alerts fire at once.

Improved SMB and Active Directory Reliability

4.19.7 includes a series of fixes for SMB service stability in Active Directory environments, focused on keeping brief glitches from turning into service outages.

Global Settings No Longer Overwritten Automatically

Previously, the global settings of a Samba cluster could be written back automatically by each node's periodic synchronization, not just changed by an administrator. If a brief connection problem occurred while the settings were being read, the system could treat them as "not yet configured" and reset them to defaults, causing a domain-joined cluster to fall back to a workgroup, breaking AD authentication, and losing custom global parameters. From 4.19.7 on, global settings are written only when an administrator explicitly changes them, such as joining or leaving AD or adding a custom parameter. If the settings cannot be read, that synchronization round is skipped and retried in the next round.

More Reliable Domain Join

When AD is enabled on a newly created Samba cluster and the trust relationship check fails after joining the domain, the system now retries immediately within the same round, up to three times, instead of waiting passively for the next synchronization round. If the domain controller appears to be temporarily unreachable, the node does not rejoin repeatedly, which protects the trust relationships of the other nodes.

Lower Background Load on Ceph

Samba's per-minute configuration synchronization now reads the complete configuration once per round instead of querying each item repeatedly, greatly reducing the constant background queries sent to Ceph. Sensitive values in the configuration, such as the AD administrator password, remain encrypted.

Fix for Multiple VIPs

Fixed an issue where, with two or more Samba VIPs configured, the same VIP was written more than once during initialization.

Stability and Security Hardening

A More Robust Management Database

The UVS management database (MariaDB Galera Cluster) now starts correctly when there is no existing cluster state, and when the database is temporarily unavailable, the scheduled report check no longer prevents the background job queue from starting. Database synchronization across nodes has also been made more robust: if the UVS Manager database becomes inconsistent between nodes because of a hardware issue, the system repairs it automatically.

Deployment and Day-to-Day Operation Fixes

Clusters can now be created when host names include a domain (FQDN, for example uvs101.example.com). Other fixes cover the SNMP service failing to start, the Update page failing to upload update packages, the "How to mount" function of a subvolume failing to create the default user, and the daily performance monitoring (PCP) job failing repeatedly after a host name change.

Security Hardening

File uploads in the Web UI now actually filter out disallowed file extensions rather than only hinting at them. PHP is upgraded to 8.4.25, and unused packages have been upgraded or removed.

Frequently Asked Questions

▾What are the main updates in UniVirStor 4.19.7?

▾Is CVE-2025-30156 fully remediated after upgrading to 4.19.7?

▾Do my clients need to be upgraded right away?

▾What can I configure in the CephX user capability editor?

▾Will upgrading to 4.19.7 interrupt service?