github Corsinvest/cv4pve-admin v2.4.0

one hour ago

This release focuses on security and permissions: a safer sign-in, and every module shows only what the user is allowed to see (review your custom roles). In the Enterprise Edition users can also sign in with Microsoft, Google, Keycloak or their company domain. It also adds the Excel export of the lists, capacity planning in the report, quick Diagnostic profiles and, in the Enterprise Edition, the automatic shutdown of the nodes from the UPS. Underlying components have been updated.

Community Edition

Added

  • Public address: the new PublicUrl setting is the address the users open. The links of the account emails (password reset, email confirmation) and, in (EE), the callback address shown for the identity providers are built from it, not from the address of the request. Without it they follow the request, as before, and a warning at startup says so.

  • Trusted proxies: the new TrustedProxies setting lists the reverse proxies the application trusts. Only through them can the user's address be passed on, so the audit log and the sign-in limits see the real address and a client cannot fake it. Without the setting nothing changes.

  • Limit on sign-in attempts: from one address, at most 10 attempts a minute to sign in or to enter the two-factor code (setting Security:MaxAttemptsPerMinute, 0 turns it off). The domain sign-in is protected too. The AI Server applies the same limit to failed keys.

  • Metrics Exporter, token in a header: Prometheus can send the token in the Authorization: Bearer header instead of the address, so it does not appear in proxy logs. The token in the address keeps working.

  • Resend email confirmation: a user whose confirmation link has expired asks for a new one from the login page, without an administrator. The page answers the same way whether the account exists or not, and limits the requests.

  • adminctl update: pulls the latest images and recreates the services that changed; self-update now suggests it as the next step.

  • Cluster settings, SSH port: each node of a cluster has its own SSH port (default 22), next to the API port, for nodes where SSH does not listen on the standard port. Existing clusters keep port 22.

  • Check for updates: a button in the Help menu looks for a new version right away, instead of waiting for the periodic check (up to 12 hours).

  • Diagnostic — profiles: Fast, Standard and Full buttons in the settings, as in Report. Fast skips the slowest reads (backup content, snapshots, LVM-thin metadata) for a quick scan of large clusters; Full turns on every optional check (S.M.A.R.T., ZFS details, CVE lookup, OK results) for audits.

  • Diagnostic — IOWait threshold: a node spending too much CPU time waiting for disks is now reported, with its own threshold (Warning 10%, Critical 25%). Before, the check used the CPU thresholds and never fired.

  • Diagnostic — new network checks: a guest using a VLAN its bridge does not let out of the node, and a guest using a bridge missing on another node (migration or HA would fail).

  • Diagnostic — PDF cover page: the PDF report opens with a cover page, as the Excel file does: report information and a table of contents with the page number of each section, clickable to jump there. The sections also appear in the bookmarks panel of the PDF viewer.

  • Excel export of the lists: an export button in the toolbar of the lists (resources, networks, disks, snapshots, cluster and node views, tasks, AutoSnap, backups, replications and, in (EE), users, roles, tokens, audit log, system logs, UPS readings and VM performance) saves them as an Excel file. Every row matching the current filter and sort is exported, not only the ones on screen, and dates are written in local time.

  • Report, Capacity Planning: a new section with one row per guest, node and storage: what is allocated, the average and peak use, how fast each storage fills up and the days left before it is full. Nodes also show the CPUs and memory given to their running guests. It uses the performance history, so turn on the history of the sections you want in it. Off by default, on with the Full preset.

  • WebHook — custom headers: HTTP headers can be added to a webhook notification (and to AutoSnap hooks (EE)), with placeholders in their values. The %severity% placeholder is now listed in the editor.

  • Enterprise features shown in the Community Edition: the Enterprise widgets (Daily Status, Git Size) are listed in the Add widget menu of the dashboard, marked with a diamond (choosing one opens the shop instead of adding it), and appear in the module overviews with the subscription notice; the Notifications page has an Apprise tab with the notice; a new Enterprise Modules page in the Admin Area lists the modules of the Enterprise Edition. Before, these features did not appear at all.

  • Command line, user disable-2fa: turns off two-factor authentication for a user who lost both the authenticator device and the recovery codes. It is the way out when that user is the only administrator.

Changed

  • Metrics Exporter (review your alerts): an unknown or disabled cluster, or a cluster without a token, now answers Unauthorized, like a wrong token, instead of Bad request: the answer no longer tells which clusters exist. Disabled is returned only with the right token.

  • Permission editor, Read of the page data removed: for the data of a page (the jobs, status and results of AutoSnap, the backups of Backup Analytics, the replications of Replication Analytics, the scans and ignored issues of Diagnostics, the scans of Update Manager, the backups of Node Protect, the reports of System Report, the UPS devices (EE)) the editor offered a Read permission that was never checked: opening the page is what shows them. It is no longer offered; roles that already have it keep it, without effect.

  • Resources, permissions per resource on every page (review your roles): the Disks, Partitions, Snapshots and Networks pages listed every VM and node of the cluster, unlike Guests, Nodes and Storages. Now they list only the readable ones, and the network diagram draws only readable nodes, guests and storages. The SDN tab needs Read on every VM, node and storage of the cluster.

  • Node Protect, permissions per node (review your roles): the module showed the backups of every node to whoever could open its pages. Now the folder backups list only the nodes the user can read; the backup of all the nodes, the Git repository of the Enterprise Edition and the widgets need Read on every VM, node and storage of the cluster.

  • Update Manager, permissions per VM (review your roles): the module showed the update status of every guest of the cluster to whoever could open its pages. Now it lists only the guests the user can read; the scan, the report and the widget need Read on every VM, node and storage of the cluster.

  • Diagnostics, permissions per resource (review your roles): the module showed the findings of every node, guest and storage to whoever could open its pages. Now the details of a scan, and the compliance view of the Enterprise Edition, list only the findings of the resources the user can read. The counts of the scans, the report, the scan, deleting a scan, ignored issues and the widgets need Read on every VM, node and storage of the cluster.

  • Replication Analytics, permissions per VM (review your roles): the module showed the replications of every VM of the cluster to whoever could open its pages. Now the replications, the trends and the scheduled replication jobs list only the VMs the user can read (the trends also only readable nodes). The scan, deleting all data and the widgets of the module need Read on every VM, node and storage of the cluster.

  • Backup Analytics, permissions per VM (review your roles): the module showed the backups of every VM of the cluster to whoever could open its pages. Now the backups, the backups in line, the unprotected disks and the trends list only the VMs (and storages) the user can read, and a scheduled backup job only when all its guests are readable. The scan, deleting all data, the log of a whole task and the widgets of the module need Read on every VM, node and storage of the cluster.

  • AutoSnap, permissions per VM (review your roles): the module showed and acted on every VM of the cluster to whoever could open its pages. Now Status lists only the snapshots of the VMs the user can read, a job is listed only when all of its VMs are readable, and Snap, Purge Snapshots, deleting and saving a job, and deleting snapshots from Status, need the snapshot permission on every VM involved. A job on @ALL, a pool, a tag or a node, and the widgets of the module, need Read on every VM, node and storage of the cluster. The picker of the VMs offers only the readable ones.

  • Language of the profile: the language chosen in the profile is applied to the interface and to the account emails sent to the user; before, both followed the language of the server.

  • Deleting a cluster also deletes the settings of the modules and the data they collected for it (scans, results, history, UPS devices, Portal tenants, tasks and permissions limited to it). The confirmation says so.

  • Maintenance, Fix All: one confirmation that lists the steps, instead of one question per step.

  • Settings labels: "Enabled schedulation" is now "Enable scheduling" (and the Activate schedulation button is now Activate scheduling), "Number which should will keep" is now "Number to keep", and the AutoSnap notify option "Allways" is shown as "Always". Existing settings are kept.

  • Help menu: the version, edition and update check share a single line at the top.

  • Date ranges: the trends of Backup Analytics, Replication Analytics and UPS (EE), and the syslog and firewall log filters of Report, pick the period in a single field instead of two separate dates. In the Backup and Replication trends the last day of the period is now included.

  • Cluster settings — nodes: the node list is edited directly, with no edit/save step per row. The order, which sets the node tried first for the API, is changed with up/down buttons instead of drag and drop, and a note explains that the same addresses are used for SSH.

  • Diagnostic: resource links in the results (and in the Compliance view (EE)) open in a new tab, so the report stays open.

  • Container IP addresses: running containers show the address they have right now, also when it comes from DHCP, in the Resources network list, the network diagram and the IP search.

  • Resources network list: new Trunks column for guest NICs.

  • Report:

    • Containers show the IP they actually have, like in Resources.
    • VM trunk ports appear in the network sheet and in the diagram; Cluster Access reports the disabled two-factor entries.
    • Rows always come in the same order, with numbers sorted as numbers (105 before 1000), so two reports of an unchanged cluster differ only in live values.
  • WebHook, placeholders: a value with quotes or line breaks no longer breaks the message or the address of a webhook, and is no longer replaced twice.

  • WebHook, errors: when the receiver refuses a notification, the error shows its answer. The log no longer shows the full address of the receiver, which for Slack, Discord and Teams contains the secret.

  • WebHook, GET requests: the body fields are hidden, since a GET request sends no body.

  • AutoSnap (review your jobs):

    • A job that selects no VM/CT, or whose guests are on an offline node, now ends as failed and the guests are listed in the log. Before, it ended OK without taking any snapshot. With On Failure Only a notification is sent.
    • Max percentage storage is now applied: guests on a storage used above the limit (95% by default) are skipped. Before, the limit was never applied. The lowest value is now 1%.
    • Guest selection: text% matches the names starting with the text and %text the names ending with it; they were inverted. An excluded range (-200:299) no longer adds guests, and pools combine with exclusions (@all,-@pool-test).
    • Containers with a bind mount or a device mount point, and VMs with a physical disk, are skipped with a message: Proxmox VE cannot snapshot them, and the job failed on them every time.
    • The timestamp format (settings) and the label (job) are checked when saved: a snapshot name longer than 40 characters, or a format that cannot be sorted in time order, is refused with the reason. Before, the job failed when it ran.
    • Selecting guests by pool on Proxmox VE 9 needs Pool.Audit instead of Pool.Allocate.
  • Metrics Exporter (review your alerts): cv4pve_replication_failed_total is replaced by cv4pve_replication_fail_count. Proxmox VE reports the failed attempts in a row, back to 0 at the first success, so the value now goes down when replication recovers. Alert on cv4pve_replication_fail_count > 0.

  • Diagnostic, codes: the first letter of a code is now always the severity of the finding. Checks with a warning and a critical threshold report the critical level with its own C code, with the same number as the W code (WN0027 and CN0027). Some codes changed number, and a problem that was reported under two codes is now reported once. Older scans keep the codes they were saved with.

  • Report:

    • VMs and Containers have new columns with the total size of the disks, the unused disks and, for VMs with the guest agent, the guest partitions.
    • JSON output: the size and the percentage of the same measure no longer share a key (memoryUsage is the percentage, memoryUsageBytes the size; the same for disk usage), and four misspelled keys are renamed (smartData, ip, position, uris).
  • Node Protect:

    • On Linux the backup archives are readable only by the user running the application, since they contain password hashes and keys.
    • When the backup of a node fails, the old backups are kept: retention is applied only after a run where every node was backed up, so good backups are never deleted to make room for an incomplete one. A run that saved no archive leaves no empty folder.
  • Permissions (review your custom roles): the actions of these pages now have their own permission, which hides the button and is checked again when the action runs. Before, opening the page was enough to run them.

    • Backup Analytics and Replication Analytics: Scan, Delete all data and the export.
    • Diagnostic: Scan, deleting and downloading a scan, creating, editing and deleting the ignore rules.
    • Node Protect: Backup, delete and download; in (EE) also the Git Push, Reset, Sync from remote and download.
    • System Report: new report, delete and download.
    • Update Manager: Scan and download.
    • Bots: sending a message.

    The built-in roles of each module include the new permissions. Custom roles do not: their users no longer see these buttons until an administrator assigns the permissions.

  • Account emails: the emails for the password reset and the email confirmation have a new layout, and the expiry they state is the real one.

  • Remembered browser at the two-factor login (EE): it is trusted for the days set in CookieSettings:ExpireDays (30 by default), like the sign-in itself. Before it was 14 days, while the profile page said 30.

  • Backup Analytics and Replication Analytics, Check widget: a new widget starts from 30 days, as many as the data kept by default, instead of 60.

Fixed

  • Security of the pages: the application can no longer be shown inside a frame of another site (clickjacking), the browser no longer guesses the type of a file, and the full address of a page is no longer sent to other sites. On HTTPS the remembered language is protected like the sign-in.

  • Web console: the console address could be misused to send other requests to Proxmox VE with the credentials of the application. Now it serves only the console.

  • Closed sessions: a session closed by the application, for example after a password reset or from the command line, stayed valid for up to two minutes.

  • Behind a reverse proxy: the redirect to the sign-in page used the internal address of the application instead of the public one.

  • Error details: the details of an error are no longer sent to the browser in production.

  • Sign-in of a disabled account: with a wrong password the answer is the same as for an account that does not exist; before, it told that the account exists.

  • Backup, file download: a folder downloaded as tar arrived as a zip file with a .tar.zst name.

  • Diagnostic, scan detail: the findings are grouped worst first (Critical, Warning, Info); the order of the groups was ignored.

  • Trends pages (Backup Analytics, Replication Analytics and, in (EE), UPS Monitor): choosing a new period keeps the storage or node and the VMs already selected; before, the first click of the range dropped them and the charts stayed empty.

  • Grids: a group, once expanded, can be collapsed again.

  • Help menu: the red dot on the Help icon is back when a new version is available, and the menu separators are lines again. The task counter no longer covers the bell.

  • Diagnostic — settings: the node RRD time frame and PSI pressure thresholds changed from the settings page are now applied; before, the scan kept using the defaults. The Storage section no longer shows RRD and PSI fields, which the storage check never used.

  • Diagnostic — fewer false alarms and more accurate checks:

    • The scan no longer stops halfway when a storage, node or guest does not answer: the missing part is reported and the rest of the scan is completed. When backups or other lists cannot be read, the checks depending on them are skipped instead of reporting "no backup" or "nothing configured".
    • Removed or corrected checks that fired by mistake: VLAN tag on a non VLAN-aware bridge (removed), unused network ports down, start on boot / locked / protection on containers, protected backups counted as old, container bind mounts, ZFS hot spares, disks behind RAID controllers, services not installed, CVEs already fixed in the installed version, and more.
    • Checks that could never fire now work: API tokens, quorum loss on node failure, failing replication jobs, pending kernel reboot, disabled storages still in use, empty pools.
    • Orphaned disks and backups now include containers and are reported once per deleted guest, and are skipped when the account cannot see every guest.
    • Numbers in the messages no longer depend on the server's regional settings (80.9% everywhere).
    • Compliance references checked against the official texts (DORA, NIS2, ENS, BSI C5).
  • Diagnostic — ignore rules to review: the check for a VLAN tag on a non VLAN-aware bridge has been removed, and the nesting without keyctl warning is now an Info finding under a new code: ignore rules written for them no longer match anything. Rules matching numbers with a decimal comma (80,9%) must be rewritten with a dot.

  • Network diagram: a guest on several bridges is linked to all of them, guests are listed by numeric id, Open vSwitch bridges are wired to their physical ports, and storages on older network configurations get their link to the bridge.

  • Guest details: NVMe disks show their wearout like SSDs; the guest agent is recognised also when enabled with the enabled=1 form; VM trunk ports are shown; options left at their Proxmox VE default (memory, swap, CPU type, OS type, …) show that default instead of an empty or zero value.

  • WebHook — stored secrets: passwords, tokens and API keys of webhooks are stored encrypted, like the other credentials. Values saved before are kept and encrypted on the next save.

  • Metrics Exporter: metrics of objects that no longer exist (a deleted or renamed guest, a removed storage, an offline node) are no longer exported forever; guest disk and network counters restart after the guest restarts or migrates; HA node states are reported; SAS disks with a healthy S.M.A.R.T. status are no longer exported as unhealthy.

  • Bots (Telegram): VM, container and node commands answer with the error of Proxmox VE when the action is refused, instead of answering as if it had worked. The built-in migrate and node shutdown aliases now work.

  • AutoSnap, deleting a job: with On remove job remove snapshots in Proxmox VE enabled, deleting a job now removes its snapshots and shows the operation among the tasks; before, the snapshots were left in place. The job stays in the list until the removal ends.

  • Dashboard: cloning a dashboard creates a copy instead of overwriting the original; an imported dashboard belongs to the user who imports it, also when the file was exported by someone else; creating the default dashboards no longer renames the Home and Status dashboards of the other users.

  • Backup Analytics: the scan imports every backup task of the retention period; before, only the last 50 of each node were read, so the first scan and busy nodes missed the older ones.

  • Replication Analytics: the scan no longer fails on a cluster without replication jobs, and the old results expire after the retention period also when the replications have stopped. A replication that fails is now stored in the run history, with its error: before, the history kept showing the last successful run and the failure never appeared. A job that keeps failing gives one record for each try a scan finds.

  • Sign-in sessions: a sign-in now survives a restart or an update of the application: before, every restart signed everyone out, also with Remember me. When a password is changed or reset, or two-factor authentication is changed, the other browsers signed in with that account are signed out at once. After the update to this version everyone signs in once more.

  • Forgot password: the reset link is sent at most 3 times to the same address in 15 minutes, and 10 requests are accepted from the same computer. Before, anyone who knew an address could fill its mailbox with reset messages.

  • Audit and privileges: deleting a backup from the Backups tab of a guest and Schedule Now on a replication are now written to the audit log, with their outcome. Deleting storage content, which the AI Server tools do, also checks the privilege on Proxmox VE before it runs.

  • Backups of a guest: New in the Backups tab of a guest now starts a backup, with storage, mode, compression, protection and notes; before, it only showed "Not available". Changing the notes or the protection of a backup is now written to the audit log. Restore, which did nothing, is no longer shown.

  • Backup Analytics, Unprotected Guests: the page was empty when no backup job was enabled, which is when every guest is uncovered. It also ignored the jobs that select a pool and the guests excluded by a job for all guests. It now evaluates every kind of job: all guests with exclusions, pool, list of IDs, and the node a job is bound to.

  • Check widgets: the Check widget of Backup Analytics, Replication Analytics and Diagnostics showed "Module not configured!" for a cluster with the schedule off, also when scans had been made by hand. It now shows those results too.

  • Diagnostics, notification: the result of a scan is now notified also when both report attachments are off; before, nothing was sent. The message carries the number of critical, warning and info issues. Who has a notifier configured and both attachments off starts to receive it: to receive nothing, leave the list of the notifiers empty.

  • Diagnostics, Community Edition: the notice in the PDF and Excel reports that names the compliance standards of the Enterprise Edition listed 14 of them. It now lists all 18.

  • Clusters: the scheduled jobs of a cluster that was deleted or disabled kept running, and failed at every run. They are now removed when the cluster is deleted or disabled, and at every start of the application. A cluster that is enabled again gets its scheduled jobs back at once, with the schedule and the options they had. The data collected for the cluster stays in the database.

  • Add cluster: with no cluster configured, every user was shown Add cluster and taken to the page that creates one, also without the permission to open it. Button and redirect are now only for who has it.

  • Command Palette: moving through the results with the arrow keys, the selected entry went out of view past the last visible result. The list now follows the selection.

  • Grids, keyboard: the edit key (F2) opened the editor of the selected row only when the last click of the mouse had been on the name column. It now always opens it.

  • Copy buttons: after a copy the button shows a green check mark for two seconds, in place of a notification.

  • Installer: install.sh changed, or refused, a database password that contains quotes or a backslash, and reduced two spaces to one. The password is now kept as typed.

  • AutoSnap, module switch: the Enabled setting of the module now stops the scheduled jobs of the cluster; before, the jobs ran whatever its value. A job is scheduled when the module and the job are both on. What is started by hand stays available with the module off. At the first start with this version the setting is turned on, once, for the clusters that have enabled jobs, so that no scheduled snapshot stops with the update.

  • Texts: the help of the AutoSnap Search Mode said that All lists the snapshots made by hand; it lists only the snapshots made by AutoSnap, on every guest of the cluster. The permission shown as "Snapshot Rallback" is now "Snapshot Rollback".

  • AutoSnap, startup: a job whose additional data is missing or cannot be read stopped the start of the whole application. A job with no additional data is now read as such; one that cannot be read is skipped, with an error in the log that names it, and the other jobs are scheduled.

  • Bots, module switch: the Enabled setting of the module now stops the bot of the cluster; before, it had no effect and the bot followed Telegram → Enabled alone. The bot runs when both are on. At the first start with this version the setting is turned on, once, for the clusters where the Telegram bot is on.

  • AI Server, permissions (review your tokens): five tools returned more than the token is allowed to see. The list of the clusters now shows only the clusters the token can work on; tasks, replication jobs and backup jobs are shown only for the guests and the nodes the token can see, and a backup job lists only the guests the token can see. A token with limited cluster permissions sees less than before.

  • Diagnostic, ignored issues: an ignore rule hides only the finding it was made for: same resource, same check, same severity. Before, a rule for VM 10 also hid VM 100, and a rule made from an Info finding also hid the Warning and Critical ones. To make a rule wider, leave its fields empty. Existing rules keep working; some findings hidden by mistake may show again.

  • Sign-in: after the login you are sent back only to a page of this installation; before, a crafted link could send you to another site.

  • Permissions: module settings open and save only with the Edit Settings permission, also from Activate scheduling. AutoSnap Snap and Purge, stopping a task and starting the update check the permission also when they run, not only when the button is shown. The Delete button of AutoSnap Status and Results is shown only to who may delete. The AutoSnap Errors page shows only the current cluster.

  • Buttons on pages that refresh by themselves: buttons that need a permission could disappear when the page refreshed by itself, for example Scan and Test Connection on the UPS devices page (EE), or be shown to someone without the permission. They now follow the user's permissions.

  • Resources, Memory Cleanup and Unlock: the two entries of the node and guest menu are shown and run only with the Power Management permission on that node or guest, and each run is recorded in the audit log. Before, any user who could see the resource could run them. Schedule Now of a replication checks its permission also when it runs.

  • Profile, language: changing the language reloads the page in the new language; before, the page lost the connection.

  • Tasks: Maintenance Fix All and the command line no longer mark as abandoned the tasks that are still running.

  • Widgets: in the Status widget of Diagnostic and of Update Manager the Critical and In error slices are red again; before, they had no colour of their own. The thin separators of the resource cards, of the Info widgets and of the issue list of the health widgets are now drawn.

  • Small fixes: in the Cluster Usage grid the memory tooltip shows the memory (it showed the storage); the disk usage tooltip of the resource lists ends with %; the default sender address of a new installation is noreply@example.com; the AI Server settings said the Toon format saves about 55% of the tokens, it is about 45%.

  • Command Palette: the vm: and ip: filters list containers too, create snapshot offers every guest and not only the running ones, and the guest picker of the commands shows only the guests you are allowed to see.

  • Cluster selector: the search ignores upper and lower case, and a search without results shows No results found! instead of hiding the selector.

  • Add cluster: the button in the top bar is back when no cluster is configured, also on the pages that do not redirect to the cluster settings.

  • Maintenance: in the Community Edition Cleanup System Logs is shown as an Enterprise feature; before, it ran and reported zero deleted records.

  • System Report: the keyboard shortcut for a new item opens the new report dialog.

  • Bots, Chats Id: the field takes one chat ID per line; before, it was a single-line box and only one ID could be entered. A value already saved keeps working.

  • Metrics Exporter: two Prometheus reads of the same cluster at the same time no longer mix their values.

  • Installer: a PostgreSQL password containing / or & no longer stops the Linux and macOS installer, and one containing $ is no longer saved altered by the Windows installer.

  • adminctl: a value with spaces in .env is read correctly; after self-update the suggested command is the one that applies a changed compose file; the help shows the documentation address.

  • Node Protect: the paths to back up are read one per line whatever line ending the setting was saved with.

  • AI Server: the descriptions the assistant reads for eight tools now match what the tools return and accept.

  • AI Server, optional parameters: the filters and options that the tools describe as optional, such as the status, the type, the detail level and the number of tasks, had to be sent on every call, and a call without them failed. They can now be left out and take the value their description states.

  • Dashboard, permissions (review your roles): the Dashboard showed every widget of every module, on every cluster, to whoever could open it. A widget is now shown for the clusters where the user holds the permission of the widget or can open its module, and reads only those clusters. The Add widget menu and the Clusters picker list only what the user can see. In a dashboard that already exists, a widget the user cannot see stays in its place with a notice and can be removed. The default dashboards are created with only the widgets the user can see.

  • adminctl, database backup: backup db create and backup all create report a failure when the database dump fails, instead of saying the backup was created and leaving an empty file.

  • Forgot password: the reset link opens a page that asks only for the new password. A link that is incomplete, altered, expired or already used shows a page that offers to request a new one, instead of an error. Setting the password from the link also confirms the email address: before, a user with an unconfirmed address could set the password and still not sign in.

  • Two-factor login (EE): a recovery code can be used at the login; before, the link led to a page that did not exist. A wrong code can be typed again without starting over from the password, and the language of the user is applied.

  • Remember me: a user who did not come back for more than 14 days was signed out, whatever the number of days in the settings. The session now lasts as long as configured.

  • Password change: changing the password, or the two-factor setup (EE), no longer signs you out a few minutes later.

  • Email confirmation: the link shows its result also when you are not signed in; before, it sent you to the login page. An altered link shows a message instead of an error.

  • Scheduled jobs: the system account is put back in place at startup when its name had been changed by the user editor.

  • Account emails translated: the emails sent to users (email confirmation, password reset, two-factor authentication turned off) now use the language of the interface instead of always English.

  • Backup and Replication Analytics, times (review your data): the times read from the Proxmox VE logs were taken in the time zone of the server running cv4pve-admin instead of the one of the node, so with the nodes in another time zone every backup and replication was stored hours off; the start and end of backup tasks were shifted too. Records already stored stay shifted until the data of the cluster is deleted and scanned again.

  • Replication Analytics:

    • The not synced for 7+ days and 24+ hours alerts and the age bars counted every stored run instead of the last run of each job: the red alert appeared even when every replication had just synced.
    • Failed and with errors were two alerts counting the same runs: one alert remains.
    • Trends reloads the charts when the period or the node changes, not only when the VM changes.
    • A replication that cannot be imported is written in the scan task, which ends as failed, instead of only in the server log.
  • Delete all data in Backup Analytics and Replication Analytics is now recorded in the audit log.

  • System Report: a report that failed stayed waiting forever; it now ends as failed, with its error, and the failure is recorded in the audit log.

  • Diagnostic: the Excel and PDF reports list the findings from the most severe, instead of starting from Info.

  • AutoSnap, jobs: the list showed the default schedule (every 15 minutes) and its next run for every job, and an empty Last Run; it now shows the real schedule, parallelism and last run. Editing a job was not affected.

  • Updater:

    • The results of a scan disappeared after 8 hours, while the default schedule runs every 12; they now stay until the next scan.
    • A Windows guest other than Windows 10/11, or a Solaris/Other guest, failed with no message: it now says that its operating system is not supported.
    • A cancelled scan sent the Scan completed email and was recorded as successful; it now sends nothing, and the dashboard no longer counts cancelled guests as in error.
    • The Download button stays usable after a report that failed.
  • Metrics Exporter:

    • The token in the Open metrics link is encoded: a token with characters such as +, & or # gave an error.
    • When the cluster cannot be reached the endpoint answers 503 with the reason, instead of 200 with old values, so Prometheus sees the target as down.
  • Notifications with attachments: when a report (for example Diagnostic) was sent to more than one notifier, the second one received an empty or unreadable attachment and failed with no trace; every notifier now gets the full attachment, and a notifier that fails is written in the log.

  • Module menu: the current page stays highlighted in the menu of a module also when its address carries extra parameters (for example the workflow list).

  • Dashboard widgets: the widgets that update during a scan could mix two updates or miss the last one; they now always show the final numbers.

  • Resources, Networks page: VMs that were stopped, or without the guest agent, were missing from the list. Now every VM is listed, with the same data as the System Report, and a new Net Name column shows the name of the network card inside the guest.

  • Diagnostic settings: more than one section can be open at the same time.

  • System Report, report dialog: it closes with a click outside, can be resized, and more than one section can be open at the same time.

Internal

  • The sign-in session no longer carries the permissions: they are always read from the database, and the session is much smaller.

  • The profile image (EE) is scaled down by the browser, and the image library is no longer used: its known vulnerabilities and its license conditions are gone. Any image the browser can read is accepted (also GIF, BMP and AVIF), up to 10 MB.

  • Underlying components updated (Proxmox VE API library, Report and Diagnostic libraries, UI components, cache, service discovery, command line). The PDF library stays on its current version: the newer one changed license.

  • Updated a .NET cryptography library with two denial of service fixes, and the .NET web and database libraries to 10.0.14.

Enterprise Edition

Added

  • External Authentication: users can sign in with an identity provider (OpenID Connect, Keycloak, Microsoft Entra ID, Google) or with an Active Directory / LDAP domain. Providers are set in Admin Area → Security → External Authentication and take effect without a restart. Test connection and Test login check them before you enable them. A provider can link existing accounts by email (never administrators, accounts with two-factor authentication or built-in accounts). It can also create the account at the first sign-in and give roles from the user's groups; the roles follow when the groups change. The sign-in page shows the providers and a Domain list. Two-factor authentication is still asked after an external sign-in. Each domain is checked at the interval set on the provider, or at once with Check accounts. Users who left the domain or were disabled lose their roles and are signed out. An account created by the domain is turned off, and turned on again if the user comes back. Deleting a provider unlinks its users and removes the roles it gave; the accounts are kept. In the users list you see the linked providers. In the user dialog you can link or unlink logins by hand and see the last external sign-ins. Tried with Keycloak, OpenLDAP and an Active Directory without encryption; Microsoft and Google not yet with real accounts.

  • Portal, Tenant Admin: each tenant can have its own admin, who sees only its tenant and manages the users of that tenant: adds new accounts, chooses their permissions on the tenant's VMs, makes another user admin. Linking an account that already exists stays with the administrator. The tenant roles are assigned by the Portal when a user joins a tenant or becomes its admin.

  • Apprise, URL Builder: a URL Builder button next to the configuration opens the online tool that writes the address of a service. In the list of services the documentation, website, secure connection and attachment support are shown as labels with a tooltip. A new Apprise notifier is filled in with the address of the Apprise service of the installation.

  • UPS Monitor, thresholds per device: the editor of a device can set its own battery low, battery critical and runtime critical alert thresholds; an empty field uses the value of the cluster settings.

  • UPS, automatic shutdown of the nodes: on a power failure the module can shut the nodes of the cluster down. Each device has a new setting, On power failure: Nothing (alerts only), the default; Dry-run (simulate the shutdown); Shutdown (shut the nodes down). With Shutdown, when a shutdown condition is met the nodes the device covers get the shutdown command of Proxmox VE, through the API of the cluster and in the order of the settings: nothing is installed on the nodes. A command that fails is tried three times, a node already off is skipped, and the node that hosts the application is commanded last. Choosing Shutdown asks for a confirmation that names the nodes. The credentials of the cluster need the privilege to shut the nodes down: it is checked once a day and the dashboard warns when it is missing. Guests are not stopped one by one, and high availability and Ceph are not handled: the shutdown test warns when the cluster has them. Tried on test nodes with a simulated UPS, not during a real power failure: start with Dry-run. Existing devices stay on Nothing.

  • UPS, shutdown simulation: with Dry-run the module says which nodes it would shut down, in which order and why, without shutting anything down. When a shutdown condition is met the nodes are recorded in Tasks and in the audit log and a notification is sent for the power failure, the condition met, the simulated shutdown and the return of the mains. No command is sent to a node.

  • UPS, shutdown conditions and nodes: new settings, separate from the alert thresholds. A device on battery meets a condition after a time on battery (600 seconds), below a runtime (10 minutes), below a charge (30%) or when the UPS reports battery low or depleted. For each node of the cluster: the order, and with more than one UPS whether all of them or any one must be in condition. The whole cluster can be treated as one, and the node that hosts the application is kept for last: set Where this application runs.

  • UPS, Test shutdown and Simulate power failure: two buttons in the dashboard. Test shutdown shows the nodes that would be shut down now, their order, the estimated time and the warnings, and changes nothing. Simulate power failure runs the whole sequence as a test, with its entry in Tasks and a notification marked TEST. During a real power failure the dashboard shows a strip with the nodes already recorded and Close episode.

  • UPS, new permissions: Change shutdown action, which also allows arming a device, Test shutdown, Simulate power failure and Close episode. The administrator role of the module has them.

  • UPS, SNMP v3: a UPS can be read with SNMP v3 instead of v2c: a user of the device, authentication (SHA-256, SHA-384, SHA-512, SHA-1, MD5) and, if wanted, encryption (AES-128, AES-192, AES-256). In the device choose the connection SNMP v3. The passwords are stored encrypted. Existing devices stay on SNMP v2c.

  • UPS, older network cards: a new SNMP v1 connection for cards that only support it, and the older DES and 3DES encryption for SNMP v3, marked as weak in the editor. A card that cannot answer many values at once is read one value at a time, with nothing to set. Tried on simulated devices only, not on real cards.

  • UPS, NUT connection: a UPS without a network card, connected by USB or serial cable to another machine, can be read from the NUT server (Network UPS Tools) of that machine. In the device choose the connection NUT, the address and port of the server (3493 by default) and the name of the UPS. The connection only reads: it never sends a command to the UPS. Stale data from the server are treated as no reading. No user is needed on the server.

  • UPS, bypass: a UPS in bypass, where the load is fed by mains without the protection of the battery, now has the status Bypass and raises a warning. Before, it was shown as Unknown and no alert was sent. The dashboard shows a Bypass counter when a device is in it. A bypass does not start a shutdown.

  • Compliance — four new standards, for a total of 18: ACN (Italian NIS2 measures), NIS2 Implementing Regulation (EU) 2024/2690, ISO 22301 (business continuity) and BSI IT-Grundschutz (Germany). They appear in the Compliance view and in the PDF and Excel reports.

  • AutoSnap — hook test: a Test button in the hook editor sends the request right away, with sample values for the selected phase, and shows the answer.

  • Users, two-factor authentication: a 2FA column shows who uses it, and Disable 2FA turns it off for a user who lost both the authenticator device and the recovery codes. The user is informed by email. The action has its own permission, included in the built-in role: custom roles need it assigned.

  • Two-factor authentication turned off: the user receives an email, whoever turned it off.

Changed

  • Portal, tenants kept apart: the Tenants page shows each user only its tenants; creating, editing and deleting tenants and adding or removing their VMs need the new All tenants permission. The Portal Tenant Admin role no longer gives permissions on all the VMs of the cluster nor the right to create and delete tenants: review your roles. Who had that role becomes admin of the tenants it was already in. A guest now belongs to one tenant only: guests already in two tenants are listed as a warning in the Vms tab.

  • Home page: it opens with the motto "By sysadmins, for sysadmins.", has a shorter text, and its link opens the documentation instead of the product page.

  • VM Performance, permissions per VM (review your roles): the Status page listed every running VM of the cluster to whoever could open it. Now it lists only the VMs the user can read.

  • UPS dashboard, disabled devices: a disabled device is no longer hidden. Its card is grey, says Disabled and shows the values of the last reading with its time; it has its own counter and is not counted as online or offline. Before, a cluster with only disabled devices showed "No UPS devices configured".

  • Users, Reset password: sends the user a link to set a new password and also shows it, so it can be delivered another way. Before, the administrator had to know the current password of the user.

  • UPS, settings (review your settings): the settings are in four tabs: Monitoring, Alerts, Shutdown and Nodes. The three alert thresholds have new names, because they only send alerts and never shut anything down: Default Battery Low Threshold is now Battery low alert, Default Shutdown Battery Threshold is Battery critical alert, Default Shutdown Time Threshold is Runtime critical alert. Their values are kept.

  • UPS, dashboard on large screens: three, four or six cards in a row, according to the width of the screen. Each card shows what the device does on a power failure.

  • UPS, continuous reading (review your settings): the devices are read continuously instead of on a schedule: every 30 seconds, every 8 seconds while a UPS runs on battery. The Cron Schedule setting is gone, replaced by Reading interval, Reading interval on battery and Unreachable after (missed readings); Enabled schedulation is now Monitoring enabled and keeps its value. A change of the settings is applied at once. The reading history keeps every reading during a power failure and at every change of status, otherwise one a minute. Run a single instance of the application: two would both send the alerts.

  • UPS, dashboard: redesigned. Counters of the devices by status, cards ordered by how much attention they need, battery and load with a bar and a small chart of the last three hours, input and output voltage, and a notice saying for how long a device has been on battery or has not answered.

  • UPS, alerts (review your notifications): a notification is sent when the alerts of a device change, not at every scan: when the UPS goes on battery, when the charge then drops below a threshold, and so on. Before, the same alert was repeated at every scan, every 5 minutes by default. A back to normal notification is sent when no alert is active any more.

  • UPS, device profile: with auto the profile is chosen by asking the device which values it has, not from the text of its description. The profile of the manufacturer is used when the device has its values, otherwise the standard one (RFC 1628). Seven profiles that could not work are removed (Legrand, Salicru, Mitsubishi, Piller, Zigor, NetVision, Borri): a device set to one of them uses the standard profile.

  • UPS, profiles checked against the manufacturers' definitions (review your devices): every profile was compared, value by value, with the published definitions of the standard and of each manufacturer, and rewritten where it read the wrong value. Ten profiles are verified and chosen by auto: Generic RFC 1628, APC, CyberPower, Eaton / Powerware, NetAgent cards (Mach Power, Ablerex, Tecnoware), Delta, HPE / HP, MGE, Tripp Lite, Riello. Socomec, Huawei, Chloride, GE and Phoenix Contact could not be verified: they work only when picked by hand and the editor says so. Vertiv / Liebert and Gamatronic are removed. A device set to a profile that no longer exists, or renamed (Eaton / Dell / ABB, CyberPower / FSP / PowerWalker, Powerware / Exide), is read with auto. Readings and status of existing devices can change: run Test Connection on each device.

  • UPS, Test Connection: the result is shown in a dialog with the profile used and one row for each value: where it is read from and what the device answered, or Missing. A warning says which alert thresholds cannot fire because the device gives no charge or no runtime.

  • Workflow engine updated (Elsa 3.9). In the list of workflows the designer opens from the row menu (Edit), no longer by clicking the first cell.

  • Workflow, permissions (review your roles): the Clone, Resize, Convert to Template and Manage HA activities now check four new permissions (Clone, Resize disk, Convert to template, High Availability), included only in the Cluster Admin role, and are recorded in the audit log. Before they acted on Proxmox VE with no check.

  • Workflow, permissions applied (review your roles): the permissions of the module were defined but not applied: whoever could open the Workflows page could edit, publish, delete and run every workflow. The designer now shows only the actions the user holds on the cluster, and the workflow engine refuses the others. Two new permissions: Publish and retract and, in History, Cancel; the administrator role of the module has them. The pickers of the activities list only the nodes, guests and storages the user can read.

  • Workflow, page address: the open workflow or run is part of the page address, so the browser Back button and the menu link return to the list and the address can be bookmarked. The Back buttons are removed, and the heading of the list, which repeated the name of the page, is no longer shown.

Fixed

  • AI Server, queries across two tables: queries that combine two tables no longer fail when both have a field with the same name, and the results can be sorted by any chosen field. The assistant is also told how to combine tables.

  • Portal: a tenant created without a description could not be saved; the editor closed and the tenant was not there.

  • Portal, deleting a tenant: its users lose their access to the guests of the tenant and the tenant roles. Before, they kept them. Access given by another tenant on the same guest stays.

  • AutoSnap, hooks: with parallel snapshots, the lines of a hook stay in the log block of their VM/CT instead of mixing with the other guests.

  • Workflow — guest configuration: for containers, the activity that reads a guest's configuration now returns the real memory, OS type, tags and protection flag, instead of empty or zero values.

  • NodeProtect — Git credentials: the username and password changed in the Git settings are now saved; before, the change was lost. The password, like the secrets of AutoSnap hooks, is now stored encrypted; values saved before are kept and encrypted on the next save.

  • AI Server, queries: the filters of a query are now applied; before, they were ignored and every row was returned. A malformed filter or a misspelled property now returns an error instead of being skipped.

  • AI Server, queries: the query tool now tells the assistant about three more filters it can use: does not contain, does not start with, does not end with.

  • Workflow, guest actions: the power action, migrate and snapshot (create, delete, rollback, update) activities now run their action; before, they never reached Proxmox VE.

  • Users, Roles and App Tokens: clicking a row selects it and opens the editor again; the Roles tab of an app token no longer drops the connection.

  • App Tokens: editing a token no longer clears its owner. A new token opens with Is Active on, and turning it off before saving is now applied.

  • Audit log: a failed creation or change of a user or a role is recorded as failed; before, it was recorded as successful.

  • Small fixes: the photo picker of a user shows only images; the Docs link of an AutoSnap hook opens the right page; Node Protect Git shows Sync from remote; the separator line on the two-factor page of the profile is drawn.

  • Apprise: the addresses are read one per line whatever line ending the setting was saved with.

  • Permissions: resetting a user password, regenerating an app token, UPS Scan now and creating a Portal tenant now check their permission when they run, not only when the button is shown.

  • Users: the system account, used by the scheduled jobs, can no longer be edited from the user list. Effective Permissions is shown only to who can read the users. The same button of the app tokens needs the read permission on tokens.

  • Portal: removing a guest from a tenant no longer removes it from the other tenants. It also removes the permissions the tenant users were given on that guest: before, they kept their access and the Portal no longer showed it. A user who is also in another tenant that has the same guest keeps them. Adding a guest that is already in another tenant now asks for a confirmation that names that tenant.

  • AI Server: in the Enterprise Edition every request to the endpoint failed with an error, also with a valid key. Requests are now served and written to the audit log with the name of the token. The query tools see only the nodes, guests and storages the token can see.

  • Audit Logs, deletion (review your roles): deleting rows of the audit log was allowed by the permission to delete users. It now has a permission of its own, Delete audit logs, which administrators hold. A custom role that could delete audit rows only because it could delete users needs the new permission to keep doing it.

  • Node Protect, Git: Reset and Sync from remote now run as tasks, with their steps in the task list and their outcome in the audit log; before, a failure left no trace. The confirmation of Sync from remote warns that the local repository is removed. On a Windows host Reset could also fail because the repository stayed open while the Git page was shown.

  • System Logs: when the logging settings had been changed by hand, the page could show the wrong logs without a warning. It now always shows the logs of the application.

  • AI Server, ISO tools: the permissions to download and to delete an ISO image could not be given to a token, because they are in no role and the permission editor did not list them. They are now listed with the other tools in the permissions of a token. They stay out of every role: no token has them until they are given.

  • Portal: Is Admin on a tenant user and Enabled on a tenant had no effect and are no longer shown; the permissions on each guest can always be set. Adding to a tenant a user whose address belongs to an existing account now asks for a confirmation that shows the roles of that account, and the display name typed is applied to an existing account too.

  • Overviews: the Overview pages of Portal and Workflow described functions the modules do not have (policies, schedules, webhooks, calls to external services). They now say what the modules do.

  • App Tokens: a role marked as default could not be given to a token. It can now be picked like any other role.

  • Workflow, Wait For Task: the Success output was true whenever the task ended in time, also when the task had failed. It is now true only when the task ended well, and the new ExitStatus output carries the outcome. Review the workflows that read Success to know only that a task had ended.

  • VM Performance: the rates per second are now those between the last two readings; before, they were averages since the page was opened, and the totals stayed at the first reading. Two readings less than a second apart no longer end in an error, and a guest that is no longer running leaves the list.

  • Workflow, commands (review your workflows): guest power actions, migrations and snapshot activities were refused in silence, because a workflow had no user, and the workflow went on as if they had succeeded. They now run with the permissions of a user: the one who started the workflow or, for a workflow started by a trigger such as Cron, the one who last published it. A command that is refused or fails stops the workflow with the reason in its history. The node shutdown and reboot activity and the backup activity now check permissions and write the audit log like the others; the backup runs on the node where the guest is, whatever the Node field says.

  • Active sessions: sending a message to the signed-in users and forcing a logout can no longer be used from outside the application.

  • Profile, two-factor authentication:

    • Reset Authenticator turns two-factor authentication off and starts the setup again. Before, it left it on with no way to configure the app again, and the next login could not be completed.
    • Disable 2FA accepts a recovery code, as the page says.
    • A remembered browser is shown, with the days it is remembered for, and Forget works.
  • Audit log: failed logins are recorded; before, none was. Password changes, password resets and every change to the two-factor setup are recorded too.

  • Users: a built-in account can no longer be deactivated from the editor.

  • UPS:

    • Test Connection said "Connection successful!" also when the device did not answer. It now reports the failure and its reason.
    • A device that does not answer becomes Offline and raises an alert after three readings in a row without an answer (the number is in the settings); before, nothing was sent.
    • With auto the standard profile was always chosen, whatever the brand.
    • A host name can be used in Host; before, only an IP address worked.
    • A wrong profile is reported as such, instead of a reading with all values at zero.
    • An APC device on battery was shown as Online with no alert, and one on mains as Unknown: each brand now uses its own meaning of the power states. A periodic self-test of the UPS is no longer taken for a power failure.
    • Wrong values in several profiles: battery voltage ten times too high on the standard profile, APC frequency ten times too low, battery voltage, runtime and temperature swapped on CyberPower and NetAgent, and others. Firmware, serial number and output power of the standard profile read the wrong value.
    • In the device list, a row opened to show its readings stayed stuck after a scan and could no longer be closed. It now stays open and its readings update when the scan ends.
  • Workflow:

    • Opening Workflows or History turned the header and the colours of the whole application to a light blue.
    • The History page listed no runs when it was the first workflow page opened.
    • On Proxmox VE 9 the list of activities in the designer was empty.
    • Manage HA on Proxmox VE 9, where HA groups are replaced by HA rules: the group list is empty and a group gives a clear error.
    • An optional field left empty (backup notes, descriptions, target node or storage, HA group, notifiers, filter expression) made the activity fail.
    • Filter Collection failed with any expression on a field.
    • Get Cluster Replications was hidden by Get Node Replications.
  • AutoSnap, webhooks:

    • The CleanJobStart and CleanJobEnd phases never fired: they now fire on Purge Snapshots and when a job is deleted together with its snapshots, and a new webhook starts from SnapJobStart.
    • Closing the webhook editor without saving kept the changes.
  • Apprise: notifications with an Excel attachment were refused by Apprise; attachments are sent correctly also when another notifier sends the same message. The editor opens at once instead of waiting for the server.

  • AI Assistant: the Copy and Download buttons under an answer did nothing. Copy now puts the text of the answer in the clipboard and Download saves it as a Markdown file. Both wait until the answer is complete.

  • VM Performance: a guest that stops or moves no longer breaks the page, and the rates are no longer negative after a guest restart. Guests that stop now leave the list and guests that start appear in it without reopening the page; on Proxmox VE 9 each disk of a guest has its own row again, with its name.

  • Diagnostic, compliance: the findings are listed from the most severe, and the list of critical findings in the PDF says how many there are instead of calling the first five a top 5.

  • Daily status widgets (Backup Analytics, AutoSnap, System Report): they show 14 days with a bar for every day, instead of 15 with gaps.

  • Edit dialogs (users, roles, tokens, External Authentication providers, tenants and tenant users): they close with a click outside, have the same width and can be resized.

Website | Docker Hub CE | Docker Hub EE

What's Changed

  • fix(diagnostic): open resource links in a new tab by @franklupo in #357
  • fix(datagrid): let an expanded group be collapsed again by @franklupo in #358
  • feat(help-menu): check for updates, update indicator on the Help icon by @franklupo in #359
  • feat: live container IPs and trunks, update SDK 9.2.3 and Report 2.6.0 by @franklupo in #360
  • feat(diagnostic): profiles, IOWait threshold, applied settings, PDF cover page by @franklupo in #361
  • docs: align user docs and README with the code by @franklupo in #362
  • fix: webhook hardening, encrypted secrets, disposed HTTP responses by @franklupo in #363
  • feat(ui): ListEditor for inline-edited lists (cluster nodes, webhook headers) by @franklupo in #364
  • ci: declare workflow permissions and job timeouts by @franklupo in #365
  • chore: update SDK 9.2.4 and cv4pve libraries, align with their behaviour changes by @franklupo in #366
  • ci: release builds the requested tag and publishes last, CodeQL on main by @franklupo in #367
  • ci: pin third-party actions, job timeouts, docs workflow permissions by @franklupo in #368
  • chore: Radzen 12.0.5, Report 2.9.0, Api.Extension 9.2.5, AutoSnap.Api 2.2.1 by @franklupo in #370
  • feat: Excel export of the lists by @franklupo in #371
  • feat: single date range field for trends and report log filters by @franklupo in #372
  • feat(report): Capacity Planning section by @franklupo in #373
  • docs: documentation site on Astro Starlight with the cv4pve theme by @franklupo in #374
  • feat: SSH port per node in the cluster settings by @franklupo in #375
  • chore(docs): force postcss-selector-parser 7.1.6 by @franklupo in #376
  • fix: language cookie marked Secure on HTTPS requests by @franklupo in #377
  • ci: documentation check on pull requests, lighter code analysis by @franklupo in #378
  • docs(seo): titles that carry the search terms, and a social card by @franklupo in #379
  • fix: apply query where conditions, purge snapshots on AutoSnap job delete by @franklupo in #380
  • fix: dashboard clone and import, tasks abandoned while running, adminctl database backup by @franklupo in #381
  • fix(backup-analytics): import every vzdump task of the retention window by @franklupo in #382
  • fix(replication-analytics): retention counted from today, job log read only when needed by @franklupo in #383
  • fix: login redirect, settings permission, unguarded actions, language change by @franklupo in #384
  • fix(diagnostic): an ignore rule hides only the finding it was made for; optional gravity and context by @franklupo in #385
  • feat(permissions): page actions need their own permission in eight modules by @franklupo in #386
  • fix(ui): theme variables that do not exist, plain font for guest selector tokens by @franklupo in #387
  • fix: small fixes found by the docs audit by @franklupo in #388
  • fix(permissions): Memory Cleanup and Unlock check a permission and are audited by @franklupo in #389
  • docs: permissions of the page actions, scope of the module permissions by @franklupo in #390
  • fix(permissions): Schedule Now of a replication re-checks the permission when it runs by @franklupo in #391
  • fix: Command Palette containers and picker, cluster selector search, tool descriptions by @franklupo in #393
  • fix: line splitting, Metrics Exporter under concurrent scrapes, installer and adminctl by @franklupo in #394
  • chore: restore the Add cluster button, gate the system logs cleanup, drop dead code by @franklupo in #395
  • docs: Matomo statistics on the public documentation site by @franklupo in #396
  • fix(security): reset password by link, 2FA recovery, sessions and cookie lifetimes by @franklupo in #397
  • docs(ups): connection test, brand detection, alerts sent on a change by @franklupo in #398
  • docs(ups): profiles verified against the manufacturers' MIB, connection test dialog by @franklupo in #399
  • docs(ups): continuous reading, reading intervals, history by the minute, new dashboard by @franklupo in #400
  • fix(permissions): PermissionView checks the permission when its parameters change, not at every render by @franklupo in #401
  • docs(ups): NUT connection by @franklupo in #402
  • docs(ups): SNMP v3 connection by @franklupo in #403
  • fix(permissions): PermissionView asks for the user of the session, not the ambient one by @franklupo in #404
  • docs(ups): shutdown simulation, shutdown conditions, nodes, renamed alert thresholds by @franklupo in #405
  • chore(tools): Proxmox VE test nodes in containers, a single node and a cluster of three by @franklupo in #406
  • docs(ups): wording of the profiles that are not verified by @franklupo in #407
  • docs(ups): automatic shutdown of the nodes, how it works and its limits by @franklupo in #408
  • fix(replication-analytics): failed replications are stored by @franklupo in #409
  • docs(portal): removing a guest revokes the permissions, guests shared between tenants by @franklupo in #410
  • fix(ai-server): tools filter by what the token can see, token identity safe for the audit by @franklupo in #411
  • fix(commands): backup deletion and Schedule Now go through a command, storage deletion checks the privilege by @franklupo in #412
  • fix(security): limit on the forgot password requests by @franklupo in #413
  • fix(security): sessions are stored in the database by @franklupo in #414
  • fix(ai-server): optional tool parameters are optional in the schema by @franklupo in #415
  • fix(ai-server): the two ISO tool permissions can be given by @franklupo in #416
  • docs(portal): controls without effect removed, existing account confirmation by @franklupo in #417
  • fix(dashboard): widgets and clusters follow the permissions of the user by @franklupo in #418
  • fix(dashboard): default dashboards hold only the widgets the user can see by @franklupo in #419
  • feat(resources): new backup from the Backups tab of a guest, edit through a command by @franklupo in #420
  • docs: Wait For Task outcome, VM Performance rates, wrong comment on the lockout by @franklupo in #421
  • fix(diagnostics): the result is notified also with both attachments off by @franklupo in #422
  • fix(jobs): the jobs of a cluster follow its deletion, disabling and enabling by @franklupo in #423
  • fix(autosnap): the module switch stops the scheduled jobs by @franklupo in #424
  • fix(bots): the module switch stops the bot by @franklupo in #425
  • docs(modules): with the switch off the manual actions stay available by @franklupo in #426
  • fix(autosnap): a job with empty extended data no longer stops the start by @franklupo in #427
  • fix(widgets): the Check widget shows the scans made by hand by @franklupo in #429
  • fix(backup-analytics): Unprotected Guests evaluates every kind of backup job by @franklupo in #428
  • fix(ui): AutoSnap search mode help and the name of the rollback permission by @franklupo in #430
  • fix: AutoSnap start with an unreadable job, installer password, Add cluster by @franklupo in #431
  • fix(ui): Command Palette arrows, edit key in the grids, copy button by @franklupo in #432
  • fix: compliance standards list, unused attribute on the layout, dollar sign in the env file by @franklupo in #433
  • docs(ups): bypass is a status of its own, with a warning by @franklupo in #435
  • Workflow: clone, resize disk, convert to template and HA through the common commands by @franklupo in #436
  • Account emails go through the translator by @franklupo in #437
  • Failed system reports saved, replication alerts per job, delete-all audited, diagnostic worst first by @franklupo in #438
  • Updater results kept until the next scan, unsupported OS message, metrics link encoded by @franklupo in #439
  • Notifier attachments per notifier, daily status 14 days, scan errors and cancellations reported by @franklupo in #440
  • AutoSnap purge runs the webhooks, jobs grid shows the real schedule, metrics answer 503 on errors by @franklupo in #441
  • Widgets refresh through their lock without losing the last change, one updater scan per cluster by @franklupo in #442
  • Backup and replication times in UTC, logs read in the node time zone by @franklupo in #443
  • Drop the redundant updater scan guard by @franklupo in #444
  • Check widget starts from 30 days, the days of data kept by @franklupo in #445
  • AI Server: measured token savings of the output formats by @franklupo in #446
  • Pin System.Security.Cryptography.Pkcs 10.0.13 by @franklupo in #448
  • Module menu selects the page also when its URL has a query by @franklupo in #449
  • Microsoft ASP.NET Core and EF Core packages 10.0.13 by @franklupo in #450
  • chore(deps): Radzen 12.1.2, ServiceDiscovery 10.11.0, System.CommandLine 2.0.13 by @franklupo in #451
  • fix(ui): warn when the Radzen side dialog resizer is missing by @franklupo in #454
  • fix(trends): keep node, storage and VM choices when the period changes by @franklupo in #456
  • fix: decisions of the open bug list (language, cluster data, Fix All, query join, images, adminctl update) by @franklupo in #457
  • feat(account): page to request a new email confirmation link by @franklupo in #459
  • fix(ui): clearer setting labels: Enable scheduling, Number to keep, Always by @franklupo in #453
  • docs(ups): per-device alert thresholds by @franklupo in #455
  • docs: Fix All, profile language, cluster deletion, query join, adminctl update by @franklupo in #458
  • docs(changelog): entries of the fixes and features since 2026-10-07 by @franklupo in #434
  • docs(workflow): permissions applied, who a workflow acts as, page address by @franklupo in #460
  • docs: third-party notices with the real licenses and links by @franklupo in #461
  • chore: remove SixLabors.ImageSharp by @franklupo in #462
  • chore: own side dialog resizer, one key handler for the grids by @franklupo in #463
  • docs(ups): SNMP v1, DES and 3DES, one value at a time, disabled devices by @franklupo in #464
  • feat(autosnap): permissions per VM, cluster-wide widgets by @franklupo in #465
  • feat(backup-analytics): permissions per VM, cluster-wide widgets by @franklupo in #466
  • feat(replication-analytics): permissions per VM, cluster-wide widgets by @franklupo in #467
  • feat(diagnostic): permissions per resource, cluster-wide widgets by @franklupo in #468
  • feat(update-manager): permissions per VM, cluster-wide widget by @franklupo in #469
  • docs(vm-performance): permissions per VM by @franklupo in #470
  • feat(node-protect): permissions per node, cluster-wide widgets by @franklupo in #471
  • feat(resources): permissions per resource on Disks, Partitions, Snapshots, Networks by @franklupo in #472
  • feat(permissions): no Read for the data of a page by @franklupo in #473
  • feat: Enterprise features shown marked in the Community Edition by @franklupo in #475
  • docs(portal): Tenant Admin, tenants kept apart, a guest in one tenant by @franklupo in #476
  • feat(security): external authentication extension point, sign-in page and docs by @franklupo in #477
  • fix(security): harden the exposed endpoints and the request pipeline by @franklupo in #478
  • chore: update packages and clean up unused usings by @franklupo in #479
  • feat(security): public address for the email links, optional trusted proxies by @franklupo in #481
  • chore(deps): bump vite from 8.3.2 to 8.3.4 in /docs by @dependabot[bot] in #474
  • chore(deps): bump undici from 8.11.2 to 8.12.3 in /docs by @dependabot[bot] in #480
  • fix(security): limit on sign-in attempts, Metrics Exporter token, lighter session by @franklupo in #482
  • docs: align the pages with the changes after 2.3.0 by @franklupo in #483
  • fix(security): two-factor code after a domain or a provider sign-in by @franklupo in #485
  • chore(deps): Radzen.Blazor 12.1.9, CsvHelper held at 33.1.0 by @franklupo in #486
  • chore(deps): Radzen.Blazor 12.2.0 and dialog fixes by @franklupo in #487
  • fix(resources): Networks page uses the same guest network data as the System Report by @franklupo in #488
  • Bump Aspire.Hosting.AppHost from 13.6.4 to 13.6.5 by @dependabot[bot] in #489
  • chore(deps-dev): bump astro from 7.3.5 to 7.4.12 in /docs by @dependabot[bot] in #490
  • chore: release 2.4.0 by @franklupo in #491

New Contributors

Full Changelog: v2.3.0...v2.4.0

Don't miss a new cv4pve-admin release

NewReleases is sending notifications on new releases.