Proxmox Backup Server 3.3.1 → 4.2.9
57 releases · 7 breaking changes · 4 security fixes · 0 advisories fixed
4security7breaking3deprecated222feature79fix
Breaking changes
- breaking fix #6267: remove old task logs and archives again on busy systems. in 4.2.7
- breaking css: remove x-grid-row-loading class, replace it with non-blurry SVG variant from proxmox-widget-toolkit in 4.2.3
- breaking fix client metadata-mode incremental backups breaking with strict payload-offset rejections introduced in pxar 1.0.1: hold-back chunks could survive across cache-range discontinuities and produce non-monotonic PXAR_PAYLOAD_REF offsets on disk. Inject any held chunk explicitly when a discontinuity… in 4.1.12
- breaking fix #6714: docs: remove wrong --drive parameter from cli example. in 4.0.15
- breaking fix #6358: remove group note file if present on group destroy in 3.4.2
- breaking fix #4382: api: access control: remove permissions of token on deletion. in 3.3.7
- breaking fix #3336: datastore: remove group if the last snapshot is removed to avoid confusing situations where the group directory still exists and blocks re-creating a group with another owner even though the empty group was not visible in the web UI. in 3.3.5
Security changes
- security backup-manager cli: document that cert update create auth and csrf key if they are missing in 4.2.6
- security re-build with proxmox-tfa 6.0.2 to cope with both possible states of backup-eligibility for WebAuthn credentials to fix a regression with (hardware) security keys from the previous version. in 4.0.11
- security Several vulnerabilities in GRUB that could be used to bypass SecureBoot were discovered and fixed (PSA-2025-00005-1) The Documentation for SecureBoot now includes instructions to prevent using vulnerable components for booting via a revocation policy. in 3.4
- security api: datastore list: move checking if a datastore is mounted after we ensured that the user may actually access it. While this had no effect security wise, it could significantly increase the cost of this API endpoint in big setups with many datastores and many tenants that each have only access… in 3.3.3
Deprecations
- deprecated update proxmox-s3-client to 1.4.2: prevent the passive request limit from overriding the active one, and name 'limit-active-requests' as the replacement in the 'put-rate-limit' deprecation notice. in 4.2.7
- deprecated api schema/section config: add support for declaring deprecated property aliases, to allow renaming properties without showing the old name in the documentation in 4.2.3
- deprecated rest server: accept deprecated property aliases in JSON request bodies by rewriting them to the canonical name before verification and dispatch, like the CLI and query-string handling already do. in 4.2.3
Every release in between
4.2.9 Nov 3, 2026
otherdocs: installer: document the advanced network options
othersnapshot locking: improve error contexts, prefer short human readable format instead of struct-debug printing
otherdatastore: prefer standard library over custom unsafe code
otherdatastore: rephrase error messages for failed chunk insert, to direct users at running verification if required
fixfix #7746: cli: manager: show installed instead of candidate version
otherGC: fail on abnormal system time drifted between phase 1 and 2, to avoid dataloss in case of manual or automatic system clock adjustments during GC runtime
otherbackup API: reject snapshots created using future timestamps outside of configured "time skew" acceptance window
otherprune: ignore snapshots with backup timestamp in the future
otherapi: sync jobs: extend privilege checks when modifying sync job to cover pre-update configuration
featureauth: ldap/ad: reject empty passwords, which might be interpreted as unauthenticated binds if the backend server allows them
4.2.8 Oct 2, 2026
othertape: labeling: improve error message for tapes with existing data, making the need for formatting them easier to understand.
fixfix #8003: sync: pull: log skipped chunk reuse in case of failed verification. When the previous snapshot is marked as failed, the chunks referenced by it are not reused, but this fact was not logged before.
fixfix #6251: api: list tasks: skip running tasks when failed ones are requested, since a task that is still running cannot have failed.
othersync: refactor and improve ACL checks when setting up local pull sync jobs, encrypting or decrypting sync jobs and prune jobs.
featureapi: admin: add missing ACL check for GC status endpoint, correctly limiting it to users with audit or backup privs.
fixfixed index files: prevent out-of-bounds memory access in FixedIndexReader and restrict fixed index files to a chunk size of 4MB in general. The backup API only allowed creating fixed index files with such a chunk size in the first place, now this is enforced across the board.
othersync: pull: refactor handling of reusable chunks to include checking of chunk sizes, in addition to chunk digests. This prevents a misbehaving pull source from potentially creating inconsistent index files on the target.
4.2.7 Sep 23, 2026
otherdocs: warn in the FAQ that apt upgrade can leave a partial upgrade, as only apt full-upgrade can remove packages to satisfy dependencies. Also fix a typo in the pcat1 file format description.
otheropenid auth: log the full error chain if getting the authorization URL fails, to ease diagnosing connection problems.
fixfix #7823: tape: do not announce a data-in transfer for SCSI commands without a data phase, which targets such as SCST rejected with an ILLEGAL REQUEST error.
othertape: configuration: also reject drive or changer device paths already used by another entry of the same type on update. Enforce unique drive numbers within each changer on creation and update. Existing duplicates can block edits until they are resolved.
othertape: permissions:
othercheck '/tape/device/{drive}' for backup, restore and the drive list, as grants on this documented path were previously ignored.
otherrequire Datastore.Read on the datastore and Tape.Write on the media pool and drive when creating or updating tape backup jobs. Scheduled jobs run as root@pam, so job editors could otherwise get a datastore written to tape without having read access to it.
othertape restore: validate the datastore name read from tape and require the snapshot name to match the one selected for restore, so corrupt or crafted media cannot place temporary files outside of the restore directory.
others3: validate object keys against the datastore layout when refreshing the contents from the bucket, instead of using them as file paths unchecked. Directory marker objects returned by some providers are now skipped.
otherclient: pxar: harden restore and extraction against crafted archives that could write or link files outside of the target directory. Symlinks are no longer followed when entering existing directories or overwriting files, also for symlinks already in the target, and hardlink targets must resolve…
otherupdate proxmox-rest-server to 1.1.1:
otheronly abort a worker task if its whole UPID matches. Previously, users allowed to stop their own tasks could stop others' tasks through a crafted UPID.
breakingfix #6267: remove old task logs and archives again on busy systems.
otherharden authentication: mask the credential check runtime in the delay for failed logins, skip malformed cookie pairs and prefer a ticket over an API token in the same request.
otheraccept quoted and unquoted Forwarded header addresses with optional ports, use the last proxy element and reject unterminated quotes.
otherhonor zero deflate weights in Accept-Encoding, stop logging redirects and unchanged responses as failures, and log client errors as warnings.
otherrun all API authentication and permission checks through one shared code path.
deprecatedupdate proxmox-s3-client to 1.4.2: prevent the passive request limit from overriding the active one, and name 'limit-active-requests' as the replacement in the 'put-rate-limit' deprecation notice.
otherupdate proxmox-network-api to 1.0.6:
fixfix #7329: allow listing interfaces configured with the IPv6 'auto' method.
otherkeep dhcp, auto and loopback methods on unrelated updates instead of switching to manual, which could drop IPv6 settings.
otherreject non-static methods while an address or gateway is set for the same address family.
otherupdate proxmox-rate-limiter to 1.0.1: prevent frequent refill updates from stalling low-rate traffic, and fix opening the shared limiter on aarch64.
4.2.6 Sep 16, 2026
securitybackup-manager cli: document that cert update create auth and csrf key if they are missing
othernotifications: use Severity::Error for acme renewal failures
otherconfig: invalidate CachedUserInfo after ACL changes, to reduce the time it takes for ACL changes to become active
fixfix #7911: api: report: mark endpoint as protected, since it calls commands that require root privileges
otherclient: improve handling of encryption/signing key configuration and manifest signature state
otherbackup: change misleading log message in sign-only mode
othermanifest: reject unsigned manifests when a key is set
otherreuse chunks across compatible crypt modes, e.g. when transitioning from unsigned to signed
featurerestore: allow explicitly trusting unsigned backups, even if the client has a key configured
4.2.5 Aug 5, 2026
otherbackup: harden the handling of client supplied backup manifests:
otheronly accept archive names that are plain file names carrying a server side type extension. A crafted name in a manifest could previously make a sync job read or write outside of the snapshot directory, running as the unprivileged 'backup' user. Reaching this needed a manifest from a configured…
otherkeep an uploaded manifest in memory and only persist it on backup finish, checking that every archive it lists was really uploaded during that session and that the checksums match the ones computed server side. A client uploading a manifest that references archives it did not upload now gets an…
fixfix #7878: sync: push: reuse the manifest of a previous snapshot on a non-encrypting push if the source snapshot was encrypted with a matching key, restoring chunk reuse and thus avoiding needlessly long sync runs.
othersync: push: keep the sign-only crypt mode of a source archive instead of reducing it to unencrypted when pushing without server side encryption.
othersubscription: reject a subscription key issued for a different architecture than the host, as arm64 keys carry an explicit marker, so a wrong key fails fast instead of only erroring during the online check.
fixupdate to proxmox-upgrade-checks 1.1, which accepts the 7.0 kernel, tells a bookworm backport apart from a trixie build and fixes the dkms check.
otherdocs: clarify in the backup protocol description that the manifest is uploaded by the client and only persisted on backup finish.
4.2.4 Jul 29, 2026
otherdocs: document the debug symbol repository
fixdatastore: fix wrong local path used for S3 bad chunk handling during garbage collection
otherrefactor file creation/mode/ownership helpers to proxmox-product-config crate
fixfix #7642: avoid expensive user lookups on file locking by caching the backup user/group ID
otherdepend on proxmox-enterprise-support-keyring, and track its version in the package version API endpoint
fixfix #5748: docs: add catalog.pcat1 format specification
otherdocs: system requirements: document we recommend local storage
fixS3: fix #6841: allow configuring request rate limits by updating to proxmox-s3-client 1.4.1. these rate limits are split into active and passive methods, allowing separate handling of POST/PUT/DELETE and GET/HEAD request limits.
featureS3: config: allow editing the use-node-config flag that controls whether requests S3 endpoints honor the node's proxy settings or not
othersync: push: gracefully handle previous manifest signature mismatches, which can happen when enabling or disabling push-encryption on an already synced backup group
4.2.3 Jul 14, 2026
breakingcss: remove x-grid-row-loading class, replace it with non-blurry SVG variant from proxmox-widget-toolkit
featurepbs-client: add backoff log throttle, to ensure progress and similar output appears quickly initially, but does not create overly long logs
otherclient: report progress during restore
featureapi: journal: adopt proxmox-syslog-api and stream the output, making the implementation consistent with the one from Proxmox Datacenter Manager
otherui: enable the structured journal view and per-service logs, including colored output and filtering capabilities
otherui: always use arrays for 'delete' property, instead of manually converting
fixfix #5971: tape: don't warn on custom MAM attribute write failures
fixfix #7175: api: time: use timedatectl instead of /etc/timezone
fixfix #7187: report: add ethtool output for physical interfaces
featureprune jobs: schedule jobs that do not prune anything, but warn during their execution. such jobs allow testing scheduling options, but make no sense for production use.
fixfix #6691: allow search by comment in datastore content, make search case-insensitive and correctly reset content view after empty searches
otherui: datastore: disable various action tooltips for actions which cannot be triggered
otherldap: escape the user-provided user name when using it in the LDAP search filter that looks up the user DN.
otherldap sync: log which user properties change when synchronizing an existing user, instead of only reporting that the user was updated.
deprecatedapi schema/section config: add support for declaring deprecated property aliases, to allow renaming properties without showing the old name in the documentation
deprecatedrest server: accept deprecated property aliases in JSON request bodies by rewriting them to the canonical name before verification and dispatch, like the CLI and query-string handling already do.
fixfix #7690: fs: replace_file: close the temporary file before renaming or unlinking it, fixing the replacement on WORM file systems and avoiding leftover .fuse_hidden files on FUSE mounts.
otherfs: make_tmp_file: append the temporary suffix instead of replacing the file extension, keeping the original file name intact for easier debugging of leftover temporary files.
4.2.2 Jun 18, 2026
otherapi: backup: run synchronous chunk-insert operations off the asynchronous runtime's worker threads. Blocking those threads, most notably during S3 uploads that can wait up to three hours for a chunk lock, could stall the I/O and timer drivers of the entire runtime and starve other backup workers.
otherclient: backup: make the file-based backup more robust against files that cannot be accessed or that vanish while the backup is running:
fixfix #7658: skip a file and log a warning instead of aborting the whole backup when querying its metadata fails with a permission error, matching the existing handling of permission errors when opening files. Such files can also still be excluded explicitly.
otherconsistently ignore files that disappear during the backup and warn about them, instead of treating this as a fatal error in some cases.
fixtape: backup: fix the command-line group filter, which was passed to the API under the wrong parameter name and thus had no effect.
otherapi: do not log a spurious error when listing the files of a snapshot whose manifest does not exist yet, which is expected while a backup to that snapshot is still running.
fixui: datastore summary: fix the per-datastore sync and prune job counts, which were derived from an incorrectly parsed datastore ID; the prune count in particular was always shown as zero.
fixtape: fix two typos in log and informational messages.
4.2.1 May 26, 2026
fixfix #5076: api: support an 'audiences' property on OpenID realms, listing additional trusted audience values besides the configured client-id. Improves compatibility with providers that issue tokens with multiple audiences.
fixfix #7562: api/ui: tape: separate the format-media 'load-barcode' parameter from the existing 'label-text' verification, so the web interface can load and format empty or previously-unrelated tapes from a changer slot in one step. The old shared parameter aborted formatting after a successful load…
othersync: pull: refuse to overwrite a locally encrypted snapshot from an unencrypted source or one using a different key, and detect content differences between two unencrypted snapshots that share a backup time. Previously such mismatches silently triggered a resync that overwrote the local snapshot.
fixdatastore: fix tuning option changes not propagating the updated sync level to the chunk store until the service was restarted.
otherdatastore: improve the error message when prune cannot acquire a snapshot lock for deletion, by showing the snapshot directory and lock file paths instead of an internal debug dump.
fixapi: backup: fix benchmark, finish-failed, and backup-failed cleanups leaving an orphaned empty backup group behind on the datastore.
otherapi: node: tasks status: return the task end time as an optional field once the task is finished, so the task viewer can render the correct duration without an extra API call.
featureapi/ui: node: add a 'location' property to the node config, exposed through the node options panel.
othersubscription: reuse the server ID from an existing subscription info when multiple candidates are detected, falling back to the first candidate only when no prior info exists.
otherchunk store: include the filesystem backing path in the syncfs log output for easier identification of which directory is being synced.
othertree-wide: print error chains as a colon-separated single line and stop quoting simple errors without context, making logs and CLI output easier to read at a glance.
otherui: encryption keys: extract the error message from API2 responses before displaying the alert, so permission and other endpoint errors no longer surface as an empty popup.
otherremovable device attach service: clarify the systemd unit description so it no longer suggests that mount is attempted for every device or that every device is a datastore.
otherdocs: faq: correct the Debian Bookworm end-of-life date; the last point release is planned for July 2026.
featureLatest 7.0 Kernel as new stable default
featureAllow moving groups and namespaces. Backup groups and namespaces can now be moved to different locations within the same datastore. This allows users to reorganize existing backups while ensuring data consistency.
featureServer-side encryption for push sync jobs. Push sync jobs can now be configured to encrypt snapshots before sending them to remote datastores. For example, this can be used to push snapshots to Proxmox Backup Server instances in less trusted offsite locations. Furthermore, pull sync jobs can be…
featureConcurrent group pull/push processing for sync jobs. Sync jobs can now process multiple groups in parallel by configuring the new worker-threads property. This can significantly improve throughput on high latency networks.
featureS3-compatible object stores as backup storage backend officially supported, including request statistics and notification thresholds. Using S3-compatible object stores as a backup storage backend has graduated from the technology preview phase and is now officially supported. S3-backed datastores…
featureKernel 7.0 as the new stable default.
featurePersist the column state of the datastore usage panel on the dashboard (issue 7412).
featureExpose new option to use the node's HTTP proxy for remote connections in the remote edit window.
featureFix an issue where creating a new user would change the pre-selected realm for the currently logged-in user (issue 7149).
featureFix an issue where OpenID Connect logins would drop the originally requested URL, preventing deep links from working correctly (issue 7337).
featureSet datastore usage related titles based on the backend type. For example, the "Storage Usage" column is now labeled "Local Cache Usage" for S3-backed datastores.
featureUpdated translations, among others:
featureCzech
featureFrench
featureGerman
featureItalian
featureJapanese
featureKorean
featurePolish
featureRussian
featureSimplified Chinese
featureSpanish
featureSwedish
featureTraditional Chinese
featureAllow moving groups and namespaces (issue 6195). Backup groups and namespaces can now be moved to different locations within the same datastore. This lets users reorganize existing backups while ensuring data consistency through per-group locking. Any lock conflicts are gracefully handled by…
featureConcurrent group pull/push processing for sync jobs (issue 4182). Sync jobs can now process multiple groups in parallel by configuring the new worker-threads property. This can significantly improve throughput on high latency networks, overcoming HTTP/2 connection limitations. Log messages are now…
featureMake the garbage collector run automatically on removable datastores when they are unmounted, if the new gc-on-unmount flag is set (issue 7465).
featureUse proper errors for fixed-index inputs instead of panics so a corrupted or unexpected index state fails the single backup task instead of crashing the service.
featureFix a bug that allowed users with powerful Sys.Modify privileges to inject arbitrary options into an apt-get invocation (PSA-2026-00003-1).
featureFix an issue which caused the API to leak information about whether a user exists if certain conditions are met (PSA-2026-00008-1).
featureSupport configuring the proxy daemon's real peer-IP header and allowed source IP addresses through environment variables (PSA-2026-00009-1). The PROXY_REAL_IP_HEADER and PROXY_REAL_IP_ALLOW_FROM variables can be set by overriding the proxmox-backup-proxy.service systemd unit. This enables…
featureAllow remote connections to optionally use the node's configured HTTP proxy (issue 3723). When configuring a remote, there is now an option to route the connection through the node's HTTP proxy.
featureImplement datastore config caching to avoid re-parsing the whole config on every access, improving performance with a large number of datastores (issue 6049).
featureFix include/exclude pattern matching for nested paths on both server and client (issue 7382). Anchored patterns in subdirectories would fail to match because the leading slash was not added when building the nested pattern.
featureFix an issue where modifying the IPv6 network configuration would overwrite an existing config method like auto with manual (issue 7329).
featureFix push sync jobs cleanup on datastore destroy.
featureFix a bug that prevented push sync jobs from being enumerated in the 'matcher-field-values' API endpoint.
featureIntegrate upstream fixes for the "Crackarmor" AppArmor vulnerabilities (PSA-2026-00010-1).
featureFix a bug where corrupted or missing job state files would prevent users from viewing configured jobs (issue 7400). The API now gracefully handles such errors by returning an unknown job state instead of failing. Furthermore, the proxy daemon now actively self-heals corrupted state files by…
featureImprove detection of configured Proxmox repositories by verifying their signature rather than relying solely on the repository URI (issue 5207). This fixes a spurious "No repository is enabled" warning in the Repositories panel when using a mirrored repository (for example, via the Proxmox Offline…
+31 more items in the release notes4.1.13 Apr 28, 2026
fixfix #5247: client: sanitize relative paths in pxar exclude patterns. Patterns starting with ./ or containing /../ silently never matched any files during a backup; they now go through the same sanitization as the paths they are matched against, and emit a warning if rewritten.
othersync: rename the internal manifest field 'change-detection-fingerprint' to 'sync-source-signature'. Avoids confusion with the unrelated client 'change-detection-mode' parameter, especially in log output. Read access falls back to the legacy field name for already-synced snapshots.
fixpull sync: fix subsequent decrypting pulls re-downloading an unchanged snapshot for the chain signed backup -> encrypting push -> decrypting pull. Decrypt-pull strips the 'sign-only' crypt-mode marker on the local target so the source-side signature check no longer reproduces; now also accept the…
otherpull sync: store and check a sync-source-signature on decrypt-on-pull, so subsequent pulls of an unchanged source snapshot can be skipped without re-downloading and re-decrypting the snapshot's files.
otherclient: pxar: rework the metadata-mode incremental-backup fix from 4.1.12 to avoid metadata-archive index bloat. Use a tighter chunk-identity check at the chunk-reuse decision sites instead of forcing injection on every cache-range hole; preserves the PXAR_PAYLOAD_REF monotonicity guarantee…
otherdatastore: dynamic-index reader: read ctime from the index header instead of the current time. Accidentally regressed by a 2020 refactor; reachable only via the proxmox-backup-debug inspect tool, which now shows the actual creation time of an inspected .didx file.
otherproxmox-backup-debug: produce deterministic output when inspecting index files (sort the chunk-digest list).
featureui: datastore content: reliably refresh the view after namespace operations. Always reload after a partial namespace move so successfully-moved groups disappear from the source view, and fix double-click navigation into a namespace that was just added by a manual reload.
featureui: datastore content: widen the actions column so all icons (including the file-browser icon) remain visible after the namespace/group move feature added another icon.
otherdatastore: when a backup group lock cannot be acquired, log the on-disk paths of the group and lock file instead of an internal BackupGroup struct debug dump.
otherdatastore: move group / move namespace: polish operator-facing log and error output. Use human-readable snapshot timestamps (not raw epoch seconds) in overlap errors, emit the "moving group X" context log line before the pre-move check so subsequent warnings have visible context, use "snapshot…
4.1.12 Apr 28, 2026
otherpull sync: surface CRC errors on the source manifest as actual errors instead of silently treating them as a vanished snapshot.
otherpull sync: bail with a snapshot-prefixed 'archive missing on source' error when fetching a referenced archive returns no data, instead of silently continuing and failing later in the index parser with a confusing message.
otherpull sync: handle vanished snapshots uniformly between local and remote sources, so a concurrent prune during a local-source sync no longer aborts the whole group with 'No such file or directory' but skips the affected snapshot like the remote path already did.
otherpull sync: clean up leftover .tmp files and the otherwise empty freshly-created snapshot directory when a source snapshot vanishes mid-sync or a referenced archive is missing, and drop the misleading 'sync done' log line for snapshots that never actually synced.
otherpull sync: emit an explicit 'skipped' log line for snapshots that fail the verified-only or encrypted-only post-filter check, so the encrypted-only rejection case is no longer silent.
breakingfix client metadata-mode incremental backups breaking with strict payload-offset rejections introduced in pxar 1.0.1: hold-back chunks could survive across cache-range discontinuities and produce non-monotonic PXAR_PAYLOAD_REF offsets on disk. Inject any held chunk explicitly when a discontinuity…
otherclient: show the full error context on upload failures in single-line form, restoring information that was lost when the metadata and payload upload streams were split.
otherui: sync/verify view: show the correct duration of finished tasks in the task log window, instead of computing the duration against the current time.
4.1.11 Apr 25, 2026
featurepull sync: avoid double-decoding blob files on the decrypt-on-pull path, which broke any pull sync job that should have decrypted snapshot manifests, catalogs or other blob files using a configured decryption key. The decrypt-on-pull feature was introduced in 4.1.9.
featurepull sync: also fetch the client log for local sync jobs, where the implementation had been a no-op stub since the trait was first introduced. Skip silently if the source snapshot has no client log, matching the existing behavior of remote sync jobs.
otherpull sync: decrypt the client log on the fly when a matching decryption key is configured, mirroring the existing decrypt-on-pull behavior for the rest of a snapshot's data.
featureui: encryption keys: always enable the tape key restore button and rename it to 'Restore Tape Key'. Restoring loads a new key, so an existing tape key does not need to be selected first; this matches the existing behavior in the tape backup view.
4.1.10 Apr 24, 2026
featureapi: add new '/nodes/localhost/identity' endpoint that returns a stable, unique instance ID derived from /etc/machine-id, for uniquely identifying a PBS node. Primarily useful for matching PBS storages in a PVE cluster to PBS remotes in PDM.
featureclient, manager: add 'server-identity' CLI subcommand to query the instance ID from the backup client and admin tools.
featurecli: migrate the 'move-group' and 'move-namespace' commands introduced in 4.1.9 from the manager to the backup client, exposed as 'group move' and 'namespace move' respectively, where they fit better alongside the other group and namespace subcommands. This also implicitly fixes any permission…
otherapi: namespace move: create the target namespace automatically if it does not exist yet (its parent namespace must still exist), matching what a plain file 'mv' does.
fixfix: sync: pull: distinguish signed-only from fully encrypted snapshots when deciding whether to decrypt on pull. The mere presence of a key fingerprint on the source manifest was previously taken as proof of encryption, which could cause pull sync jobs with a matching decryption key to fail on…
otherpartially fix #6372: acme: also pick a shorter renewal lead of half the certificate lifetime for short-lived certificates (under ten days total), following the Let's Encrypt integration guide. The previous 1/3 fraction still applies for longer-lived certificates.
otherui: datastore create: drop the now outdated 'technology preview' label for S3-backed datastores.
otherui: tape: tune vertical alignment of the tape icon in tree views so it sits better next to the surrounding text.
otherdocs: describe the S3 request counters, how they can trigger notifications, and the periodic threshold reset schedule.
otherdocs: list previously missing entries under the configuration subtree with a short description of each.
4.1.9 Apr 23, 2026
others3: graduate the S3 datastore backend from technology preview to a fully supported feature; drop the corresponding preview note from the docs.
fixfix #7251: sync jobs: support server-side encryption of snapshots using encryption keys stored on the datastore server. Push sync jobs can encrypt previously unencrypted source snapshots on the fly while uploading to the remote; pull sync jobs can decrypt snapshots with a matching key and store…
featureconfig: add 'encryption-keys' as ACL subpath under 'system' for more granular access control over individual encryption keys.
featureui: expose encryption key management via a new menu (list, create, archive/unarchive and delete keys) and allow assigning an active encryption key and associated decryption keys in the sync job editors.
fixfix: sync: pull: also upload the client log to s3-backed target datastores in the "no data changes" fast-path; this was previously only done on full resync.
otherdocs: add chapter on server-side encryption for sync jobs, covering key management, associated keys and requirements for removing keys.
otherui: sync job edit: show a direction-specific hint in the encryption panel, since pull sync jobs only use configured keys to decrypt and never encrypt, which is easy to miss without an explicit note.
otherapi/ui: sync job: reject setting the active encryption key on pull sync jobs (including switching a push job with an active key over to pull without clearing it first), since the active key is only consumed by the push side and would otherwise be silently ignored.
featureadd move operations for backup groups and entire namespace subtrees within the same datastore, covering both filesystem and S3 backends and preserving group notes. Snapshots can be merged into an existing target group when ownership matches and source snapshots are strictly newer. Exposed via new…
featuredatastore: add a per-datastore move journal consumed by the mark phase of garbage collection, closing a race where a concurrent move-group or move-namespace could cause the mark phase to miss a relocated index, which would let the sweep phase remove its still-referenced chunks.
otherdocs: add section on moving namespaces and groups.
otherui: show empty backup groups. Previously, removing the last snapshot of a group also removed the group itself, discarding group-level metadata like notes; it also prevented cleaning up groups left behind by a partially failed move on an S3-backed datastore without going through the API.
otherpartially fix #6372: acme: scale certificate renewal check cadence with certificate lifetime. Start renewing once 2/3 of the total lifetime have passed instead of the hard-coded 30 days, with a minimum of 3 days remaining to still cope with transient failures. This lets short-lived (e.g. 7-day…
otherdocs: sync: document the new 'worker-threads' option for parallel sync jobs.
otherdocs: storage: document the 'gc-on-unmount' flag for removable datastores.
4.1.8 Apr 22, 2026
fixfix #4182: sync jobs: allow pulling or pushing backup groups in parallel, configurable via the new 'worker-threads' sync job property (1 to 32, defaults to 1 for sequential). Significantly improves throughput on high-latency connections where the single HTTP/2 connection to the source was limited…
othersync jobs: in parallel mode prefix log messages by group and buffer them briefly, so lines from concurrent groups stay readable and do not interleave excessively.
featureui: sync job edit: expose the new parallel group worker count setting.
featureapi: datastore: add 'gc-on-unmount' option for removable datastores, which runs garbage collection automatically before unmounting. This gives setups relying on auto-unmount a natural point to run GC. The unmount waits for the GC run, or an already running one, to complete. Rejected on…
featureui: datastore options: expose the new 'gc-on-unmount' toggle for removable datastores.
featuretools: move ParallelHandler to the new proxmox-parallel-handler crate in the proxmox-rs workspace, for reuse across Proxmox products.
otherbump proxmox-shared-memory to 1.0.2, which immediately flushes the contents of newly created mmap files, avoiding a possible race between a file being initialized and another consumer opening it while checking its header.
featureclient: add --server, --port, --datastore, --auth-id and --ns as per-field alternatives to the compound --repository URL across all client tools (proxmox-backup-client, proxmox-file-restore, proxmox-backup-debug). Each has a matching PBS_SERVER, PBS_PORT, PBS_DATASTORE, PBS_AUTH_ID environment…
fixfix #5340: client: add PBS_NAMESPACE environment variable as a per-tool fallback for --ns, consistent with the other PBS_* atom environment variables.
otherdocs: document the new per-field client repository options, their mutual exclusion with --repository, and the corresponding PBS_* environment variables.
fixfix #7337: ui: preserve the target deep link when redirecting through an OpenID login round-trip. The hash was previously dropped, sending users to the dashboard instead of their originally requested view after logging in.
fixfix #7400: handle corrupted or missing job statefiles gracefully. The API no longer errors out when the statefile cannot be parsed (a default schedule status is returned instead), and the scheduler loops actively overwrite corrupted statefiles on the next run so affected jobs resume normally.
fixpbs-key-config: fsync() newly stored encryption keys to match the replace path, so a crash between key creation and the next filesystem flush cannot leave the key partially written on disk.
4.1.7 Apr 15, 2026
otherbump pxar dependency, adding a consistency check for payload reference offsets in split archives.
fixfix #7382: correctly anchor nested paths for include/exclude patterns. Anchored patterns in subdirectories were not matched because the leading slash was not prepended to the match path.
fixfix #7329: api: avoid overwriting existing IPv6 config method. Setting IPv4 configuration could silently reset the IPv6 method to manual, causing the inet6 stanza to be dropped from the interfaces file.
othersystem report: adapt to the ZFS arcstat CLI tool getting renamed to zarcstat.
fixfix #7412: ui: dashboard: persist datastore usage column configuration across page reloads and navigations.
otherui: task descriptions: use title case also for 'Format Media' task.
others3: bump proxmox-s3-client to 1.4.0, fixing the Content-Type for object upload and copy requests to application/octet-stream instead of the non-standard binary/octet.
fixfix #6716: pass node http proxy config to s3 backend, so that S3 operations respect the configured HTTP proxy.
features3: add request statistics with configurable notification thresholds and scheduled counter resets, allowing to monitor and alert on S3 backend traffic.
otherpartially fix #6563: ui: expose s3 rrd charts in datastore summary.
otherapi: metrics: support sub-path metric object IDs, so that S3 sub-metrics inherit their parent datastore's permissions.
otherapi: datastore: include backend type in listing response, allowing the UI to distinguish backends without loading the full config.
otherui: s3: hide S3-only options for non-S3 datastores.
otherui: datastore summary improvements:
othergive notes panel its own full row.
otheruse backend-type from datastore list API for immediate S3 detection.
4.1.6 Mar 19, 2026
fixfix #3723: api: remote: optionally use node proxy config for http client
fixfix #7078: ui: S3: expose "DeleteObjects via DeleteObject" provider quirk
otherapi: status/datastore: expose backend type in datastore status/list
otherui: Set datastore usage related titles based on backend
otherbump proxmox-fuse to 3
fixfix #7017: cache token.shadow verification results
4.1.5 Mar 10, 2026
fixapi: backup: don't verify total file size for incremental backups to fix a critical regression with the previous version bump and incremental qemu backups. As there, the submitted value is currently the upload size instead of the total size.
otherapi: backup: disallow incremental backups with different index length again. Currently, fixed index files are only reused for qemu backups where the total size never changes at the moment.
4.1.4 Feb 17, 2026
fixupdate proxmox-rest-server to 1.0.5 addressing a minor regression that resulted in some requests with auth issues being answered with HTTP code 400 "bad request" instead of the correct HTTP code 401 "unauthorized". The new rest-server also defaults to only checking the forwarded header in the…
featureproxy daemon: support configuration of real peer-ip header and allowed source through the PROXY_REAL_IP_HEADER and PROXY_REAL_IP_ALLOW_FROM environment variables. These variables can currently be set through overriding the proxmox-backup-proxy.service systemd unit. For example, to check the…
4.1.2 Jan 29, 2026
fixfix #7149: ui: user create: do not leak changing the realm for a new user into the pre-selected realm of the currently logged in user, as that might cause some UX friction and confusion on their next login.
fixfix recent regression that failed s3 refresh.
fixfix #7219: client: mount: align encryption key loading behavior to fallback to the default key location, just like all other commands.
featurepartially fix #6049: datastore: implement various optimizations for handling large datastore configs, including a cache for the datastore config to avoid re-parsing the whole config if not necessary.
otherchunk store: various code clean-ups for the chunk store iterator, chunk filename checks and "bad chunk" handling.
featureadd basic multiarch support to resulting package and executable locations.
featureswitch ACME implementation over to factored out one.
fixfix #6939: acme: support servers returning 204 for nonce requests.
4.1.1 Dec 22, 2025
otherdatastore: s3: use owner from local store cache instead of refetching to reduce the time and amount of request needed for listing backup contents.
featuregarbage collection: add task-cancellation points during phase 2 for s3 stores after each processed s3 object key.
othertape: cope with drives that aligned buffer sizes when setting up encryption.
otherapi: notification: correctly enumerate known push-sync-jobs.
fixapi: config: fix push sync jobs cleanup on datastore destroy.
otherupdate proxmox-apt and related api-types for more robust package name handling.
featureLatest 6.17.2-1 Kernel as new stable default
featureTraffic control based on the authenticated user. Traffic control already allows limiting backup and restore traffic for specific client networks. Now, traffic control can also take the authenticated Proxmox Backup Server user into account. This allows more convenient and fine-grained…
featureAllow to control parallelism of verify jobs. Backup snapshot verification is both an IO- and CPU-heavy workload, as it needs to read chunks from disk and verify their checksums. The number of threads used for reading from disk and for checksum verification can now be customized. In some setups,…
featureBandwidth rate limiting for S3 endpoints. Since Proxmox Backup Server 4.0, backup metadata and chunks can be stored in S3-compatible object stores (technology preview). It is now possible to enforce bandwidth rate limits on incoming and outgoing connections from or to S3 endpoints. This can avoid…
featureKernel 6.17 as the new stable default.
featureAllow users outside the pam realm to access the shell (issue 6398). Accessing the shell requires the Sys.Console privilege. Non-PAM users are required to enter valid PAM credentials to access the shell.
featureFix a sorting issue where the "Last comment" field was not considered (issue 6592).
featureImprove appearance with high-resolution displays.
featureFix an issue where the WebAuthn setting dialog, which is only accessible for the root user, was susceptible to XSS (PSA-2025-00014-1).
featureUpdated translations, among others:
featureCzech
featureFrench
featureGeorgian
featureGerman
featureItalian
featureJapanese
featureKorean
featurePolish
featureSpanish
featureSwedish
featureTraditional Chinese
featureUkrainian
featureAllow to control parallelism of verify jobs. Backup snapshot verification is both an IO- and CPU-heavy workload, as it needs to read chunks from disk and verify their checksums. The number of threads used for reading from disk and for checksum verification can now be customized in the datastore…
featureAllow to automatically unmount removable datastores after an automatic sync job is completed. A removable datastore can be configured to trigger a sync job when it is mounted. In addition, the removable datastore can now be unmounted automatically after the sync job has finished.
featureAllow removable datastore creation on the device's root directory.
featureImprove performance of the API endpoint for listing backup groups.
featureEnsure proper cleanup if setting the backup group owner fails.
featureClarify task log messages for protected backup snapshots when pruning (issue 6624).
featureMount a dedicated tmpfs under /run/proxmox-backup to avoid exceeding the maximum number of inodes on low-memory hosts. This will only take effect after a reboot.
featureFix an issue where Proxmox Backup Server would stop accepting incoming connections in certain error cases. This could be caused by health checks, for example when running behind HAProxy.
featureFix a regression where pull sync jobs would not resync all corrupt snapshots.
featureFix an issue where authenticating against a Remote running some versions of Proxmox Backup Server 3.4 would fail.
featureFix an issue where backup snapshot creation would not drop locks in time and cause an immediately following backup snapshot creation to fail (issue 6566). This could, for example, happen for push sync jobs.
featureFix a race condition during long-running garbage collection and pruning of recent snapshots that could lead to backup corruption before Proxmox Backup Server 3.4 (PSA-2025-00019-1).
featureImprove the efficiency of fingerprint validation.
featureFix a verification error message that mentioned S3 even in case of a local datastores.
featureFix a locking issue where a backup would not be prioritized over garbage collection during a server reload.
featureAvoid slowdowns when waiting for chunk insertion to finish during pull sync jobs.
featureEnsure correct reuse of pre-existing chunks for backup snapshots created by a pull sync job.
featureFix a race condition between deleting a datastore and modifying the configuration of unrelated datastores.
+42 more items in the release notes4.0.22 Nov 24, 2025
fixs3 garbage-collection: fix local marker cleanup for unreferenced chunks that are only available on the s3 storage itself.
otherprovide and enable dedicated mount unit for /run/proxmox-backup, avoiding the (inode) limits of /run.
featuredatastore config: add default-verification-workers and default-verification-readers to the tuning options.
featureui, docs: integrate new verification related thread-count options.
featureui: traffic-control: rename add-timeframe button to "Add Timeframe" to avoid having two buttons with an "Add" label visible at the same time.
4.0.21 Nov 21, 2025
fixtask tracking: fix adding new entry if other PID is tracked.
otherGC: S3: reduce number of open FDs for to-be-deleted objects to avoid running into soft FD count limits, as we cannot just increase that due to some libraries still potentially using the ancient select syscall.
othersystem report: better hardware and system utilization overview.
otherpulling sync: ensure that all chunks of a synced index files can be downloaded from the remote.
otherpulling sync: mark chunks referenced in last backup snapshot as reusable.
otherpulling sync: keep one "encountered chunks" map for a full backup group to avoid frequently reinstating that map.
otherdocs: storage: fix "persistent" typo
featuretraffic-control: add user-specific rule matching and precedence. Introduce user-aware rule lookup by accepting an optional user and preferring rules that specify 'users' over IP-only matches.
featureui: traffic-control: add users field in edit form and list.
features3 datastore: add support for configuring bandwidth rate-limiting for clients, reusing the token bucket filter that the API server uses for rate limiting.
otherui: expose rate and burst limits for s3 endpoints.
4.0.20 Nov 14, 2025
otherverify: distinguish s3 object fetching and chunk loading error, as loading chunks from the object store might be cause by transient issues, and must therefore handled so they do not incorrectly mark chunks as corrupt.
otherapi: admin: rework waiting on active operations and maintenance locking.
otherapi: datastore: wait for active operations to clear before s3 refresh.
othernotifications/pbsXtoX: adapt to proxmox-apt making old_version optional.
featureapi: access: add vncticket verification endpoint.
featureapi: node shell: allow access for tokens.
otherui: replace Proxmox logo with slightly bigger SVG version to make it crisper on HiDPI displays.
otherdatastore: GC: drop overly verbose info message during s3 chunk sweep.
fixdatastore: various concurrency fixes for GC and handling corrupt chunks.
otherverify: never hold mutex lock in async scope on corrupt chunk rename to avoid rare deadlock potential.
othercleanup chunk markers from cache in phase 3 on s3 backends.
featureverification: add option to set number of {read,verify}-threads for jobs, allowing one to potentially increase verification throughput as long as the system does not gets IO starved.
featureapi: datastore: auto-unmount after mount-triggered sync for a removable datastore. This can be configured through a new 'unmount-on-done' option for sync jobs.
4.0.19 Nov 7, 2025
featuresync: pull: instantiate backend only once per sync job to avoid establishing a new connection for each chunk that gets inserted.
otherapi, datastore: ensure that the exclusive lock on the backup group is acquired before updating the notes, in order to avoid possible race conditions, e.g. with backup group destruction.
otherbackup log upload: avoid calling a blocking function in async context.
otherverify: improve logs and behavior on errors.
otherdatastore: avoid potential loss of concurrent config changes that happened during datastore deletion.
fixfix #4995: pxar: Include symlinks in zip file creation.
fixchunk store: fix theoretical race window between chunk stat and GC cleanup, for a case where a chunk must be older than the cutoff time and not be referenced by any index file and also uploaded at the exact same time that the GC passed over it in phase 2.
othervarious code style fixes, partially automated through rust clippy, upgrading from rust 1.85 to 1.90..
4.0.18 Oct 22, 2025
otherapi: chunk upload: extend consistency checks for s3 backend
4.0.17 Oct 15, 2025
fixfix #6665: never rename chunks on s3 client fetch errors
fixfix #6398: api: allow non-pam users to access shell
otherapi: admin: datastore: optimize groups api call
othersync job: pull: avoid blocking future during thread pool completion
otherrefactor GC helpers
otherS3 chunk cache: rework fetching and caching logic
fixfix #6906: s3-client: conditionally set Content-Length header
otherGC: refactor chunk removal helper
4.0.16 Oct 2, 2025
fixsync: pull: fix regression in resync of corrupt snapshots
fixfix #6750: api: avoid possible deadlock on datastores with s3 backend
fixfix #6566: backup: api: conditionally drop group and snapshot locks
otherGC: rework locking logic to avoid deadlock with s3 backend
otherverify: silence incorrect backend error message for filesystem stores
4.0.15 Sep 15, 2025
otherpbs3to4: sync to changes made on bookworm PBS 3 branch.
othernotification: don't deserialize notification config when there is nothing to send.
otherapi: do not block current thread during s3 backup log upload.
others3: pass now optional default request timeout for s3 api calls.
breakingfix #6714: docs: remove wrong --drive parameter from cli example.
fixfix #6624: ui: improve task log output for protected entry.
featureupdate proxmox-s3-client to 1.2.2 to benefit from the higher request timeout of 30 minutes and the new retry logic for transient client errors using an exponentially increasing back-off time and a maximum of 3 tries for the retries.
featureapi: allow creation of removable datastore at device's root.
otherdocs: network-management: use correct syntax for code-blocks.
otherdocs: network-management: link to systemd.link documentation.
fixfix #6592: ui: add custom sort function to comment column.
4.0.14 Aug 13, 2025
otherui: webauthn view: make sure renderers are html encoded
4.0.13 Aug 11, 2025
otherbin: pbs3to4: adapt boot-loader checks to trixie
otherdatastore: cleanup newly created group dir if setting owner failed
4.0.12 Aug 7, 2025
otherupdate proxmox-tfa dependency to 0.6.3 to also keep old passkeys created with PBS 3 or earlier working, where the backup eligible state was not yet exposed.
featureapi: openid connect: allow users of OIDC realms to opt into the HttpOnly cookies.
featureui: opt OpenID connect authentication flows into the new http only flow.
otherproxy: avoid exiting connection acceptor loop in error case, like some monitoring and health checks of reverse proxies or load balancer cause.
4.0.11 Aug 6, 2025
securityre-build with proxmox-tfa 6.0.2 to cope with both possible states of backup-eligibility for WebAuthn credentials to fix a regression with (hardware) security keys from the previous version.
4.0.10 Aug 5, 2025
featurenotifications templates: adapt to whitespace handling changes with new handlebars version to ensure lists are not placed on a single line.
featureS3 Endpoints: allow one to set quirks to adapt to provider specific behavior, like not understanding the If-None-Match HTTP header.
othermetric collection: use ip link for determining the type of interfaces.
otherd/control: recommend proxmox-network-interface-pinning package for the Proxmox Backup Server package.
otherdocs: add documentation for proxmox-network-interface-pinning.
4.0.9 Aug 4, 2025
othersync: conditionally pre-filter source list by verified or encrypted state.
otherapi: return s3 client errors during datastore create with context.
featureacls: add s3-endpoint as valid 'system' subpath and expose on the web UI.
others3 endpoint: make required permissions specific to acl sub-paths in '/system/s3-endpoint/{id}' to allowing one to provide only access to a specific S3 endpoint.
4.0.8 Jul 31, 2025
featureadd pbs-network-config-commit systemd service to apply pending network changes on boot.
otherdatastore: delete all objects on datastore destroy when the remove data flag was set.
otherui: switch datastore destroy label text depending on the backend type.
otherdocs: s3: extend recommendations and limitations for local cache.
otherdatastore: destroy: delete s3 in-use marker unconditionally to make re-adding the bucket as datastore easier.
otherdatastore: wrap bucket name, as in is now optional in the s3 client.
otherapi: admin s3: make store prefix for check command optional.
featureapi: config s3: add bucket list api endpoint.
otherui: replace bucket field by bucket selector
4.0.7 Jul 25, 2025
features3 api types: make region regex less strict to allow more providers, like e.g., Hetzner, to work.
othertell client to remove authentication cookies via http header on unauthorized request.
featureallow log-in via valid parameters even if there is a pre-existing HttpOnly cookie with a (now) invalid ticket, as that should not block valid new logins.
otherdon't set Expire for HttpOnly cookies anymore as heuristic to try to make browsers clear these login-ticket cookies on exit.
4.0.6 Jul 24, 2025
fixd/postinst: fix setting up pbs-test repo on fresh installation during the beta.
featureLatest 6.14.8-2 Kernel as new stable default
featureNew major release based on the great Debian Trixie.
featureSeamless upgrade from Proxmox Backup Server 3.4, see Upgrade from 3 to 4.
featureS3-compatible object stores as backup storage backend (technology preview). Proxmox Backup Server can now connect to object stores that provide an S3-compatible API. A new datastore type persists backup metadata and chunks to an S3-compatible object store. To reduce the number of requests to the…
featureZFS now supports adding new devices to existing RAIDZ pools with minimal downtime.
featureFix an issue where the datastore creation tasks would not be formatted properly in the task list.
featureImprovements to the custom consent window:
featureAvoid showing the consent text twice for OIDC logins.
featureRestrict the text length of the consent banner to 64 KiB.
featureFix an issue where the consent window size would exceed the viewport size (issue 6312).
featureUse HttpOnly cookies as an additional layer of protection against cookie stealing attacks.
featureSmall improvements to the services view in the host's system panel.
featureFix an issue where the node services panel would incorrectly show postfix as disabled.
featureImproved handling of translations:
featureAdd support for plural forms and ngettext usage.
featureTranslations can now contain comments that are extracted from the source code and provide useful context for translators.
featureUpdated translations, among others:
featureCzech (new!)
featureArabic
featureBulgarian
featureFrench
featureGerman
featureItalian
featureJapanese
featureKorean
featurePolish
featureRussian
featureSimplified Chinese
featureSpanish
featureSwedish
featureTraditional Chinese
featureUkrainian
featureS3-compatible object stores as backup storage backend (technology preview). Proxmox Backup Server can now connect to object stores that provide an S3-compatible API. A new datastore type persists backup metadata and chunks to an S3-compatible object store. One such datastore cannot be shared…
featureAllow to automatically run sync jobs whenever a relevant removable datastore is mounted. Sync jobs that reference a local removable datastore can only be run while the removable datastore is mounted. Setting the new run-on-mount flag on a sync job causes it to run automatically when the removable…
featureFix an issue where forgetting a group, or the last snapshot of a group, would fail if a group note is present (issue 6358).
featureFix an issue where forgetting a group, or the last snapshot of a group, would fail if the owner file was already deleted.
featureImprovements to the garbage collection:
featureAllow bypassing the chunk cache used during the marking phase by setting a cache capacity of 0.
featureTrack chunk cache statistics in the garbage collection task log.
featureFix logging in case of index files created or deleted while the garbage collection job is running.
+26 more items in the release notes3.4.9 Sep 26, 2026
3.4.8 Sep 26, 2026
3.4.7 Sep 26, 2026
3.4.6 Sep 26, 2026
3.4.5 Sep 26, 2026
3.4.4 Sep 26, 2026
3.4.3 Sep 26, 2026
3.4.2 Jun 4, 2025
otherdatastore: various small perf optimizations
otherrestore-daemon: adapt to zpool output changes in 2.3
breakingfix #6358: remove group note file if present on group destroy
otherdocs: tuning: list default and maximum values for gc-cache-capacity
otherbackup info: avoid additional stat syscall for protected check
othertools: lru cache: document limitations for cache capacity
othergarbage collection: bypass cache if gc-cache-capacity is 0
3.4.1 Apr 16, 2025
fixui: token view: fix typo in 'lose' and rephrase token regenerate dialog message for more clarity.
otherrestrict consent-banner text length to 64 KiB.
otherdocs: describe the intend for the statically linked pbs client.
otherapi: backup: include previous snapshot name in log message.
othergarbage collection: account for created/deleted index files concurrently to GC to avoid potentially confusing log messages.
fixgarbage collection: fix rare race in chunk marking phase for setups doing high frequent backups in quick succession while immediately pruning to a single backup snapshot being left over after each such backup.
othertape: wait for calibration of LTO-9 tapes in general, not just in the initial tape format procedure.
featureLatest 6.8.12-9 Kernel as new stable default
featurePerformance improvements for garbage collection. Garbage collection frees up storage space by removing unused chunks from the datastore. For this, it has to mark all chunks that are currently in use by backup snapshots. The marking phase now uses a cache to avoid redundant marking operations. This…
featureMore fine-grained control over backup snapshot selection for sync jobs. Sync jobs are useful for pushing or pulling backup snapshots to or from remote Proxmox Backup Server instances. Group filters already allow selecting which backup groups should be synchronized. In addition, it is now possible…
featureStatic build of the Proxmox Backup command-line client. Proxmox Backup Server is tightly integrated with Proxmox VE, but its command-line client can also be used outside Proxmox VE. Packages for the command-line client are already provided for hosts running Debian or Debian derivatives. A new…
featureLatest Linux 6.14 kernel available as opt-in kernel.
featureAllow configuring a default realm which will be pre-selected in the login dialog (issue 5379).
featureThe prune simulator now allows specifying schedules with both range and step size (issue 6069).
featureEnsure that the prune simulator shows kept backups in the list of backups.
featureFix an issue where the GUI would not fully load after navigating to the "Prune & GC Jobs" tab in rare cases.
featureDeleting the comment of an API token is now possible.
featureVarious smaller improvements to the GUI.
featureFix some occurrences where translatable strings were split, which made potentially useful context unavailable for translators.
featurePerformance improvements for garbage collection (issue 5331). Garbage collection frees up storage space by removing unused chunks from the datastore. For this, its first phase updates the atime (access time) of all chunk files that are currently in use by backup snapshots. The marking phase now…
featureAdd a safeguard against filesystems that do not honor atime updates (issue 5982). The first phase of garbage collection marks used chunk files by explicitly updating their atime. If the filesystem backing the chunk store does not honor such atime updates, phase two may delete chunks that are still…
featureAllow to customize the atime cutoff for garbage collection in the datastore's tuning options. The atime cutoff defaults to 24 hours and 5 minutes, as a safeguard for filesystems that do not always immediately update the atime. However, on filesystems that do immediately update the atime, this can…
featureAllow to generate a new token secret for an API token via the API and GUI (issue 3887).
featureRevert a check for known but missing chunks when creating a new backup snapshot (reverts fix for issue 5710). This check was introduced in Proxmox Backup Server 3.3 to enable clients to re-send chunks that disappeared. However, the check turned out to not scale well for large setups, as reported…
featureEnsure proper unmount if the creation of a removable datastore fails.
featureRemove a backup group if its last backup snapshot is removed (issue 3336). Previously, the empty backup group persisted with the previous owner still set. This caused issues when trying to add new snapshots with a different owner to the group.
featureDecouple the locking of backup groups, snapshots, and manifests from the underlying filesystem of the datastore (issue 3935). Lock files are now created on the tmpfs under /run instead of the datastore's backing filesystem. This can also alleviate issues concerning locking on datastores backed by…
featureEnsure that permissions of an API token are deleted when the API token is deleted (issue 4382).
featureEnsure that chunk files are inserted with the correct owner if the process is running as root.
featureFix an issue where prune jobs would not write a task log in some cases, causing the tasks to be displayed with status "Unknown".
featureWhen listing datastores, parse the configuration and check the mount status after the authorization check. This can lead to performance improvements on large setups.
featureImprove the error reporting by including more details (for example the errno) in the description.
featureEnsure that "Wipe Disk" also wipes the GPT header backup at the end of the disk (issue 5946).
featureEnsure that the task status is reported even if logging is disabled using the PBS_LOG environment variable.
featureFix an issue where proxmox-backup-manager would write log output twice.
featureFix an issue where a worker task that failed during start would not be cleaned up.
featureFix a race condition that could cause an incorrect update of the number of current tasks.
featureIncrease the locking timeout for the task index file to alleviate issues due to lock contention.
featureFix an issue where verify jobs would be too eagerly aborted if the manifest update fails.
featureFix an issue where file descriptors would not be properly closed on daemon reload.
featureFix an issue where the version of a remote Proxmox Backup Server instance was checked incorrectly.
featureStatic build of the Proxmox Backup command-line client (issue 4788). Proxmox Backup Server is tightly integrated with Proxmox VE, but its command-line client can also be used outside Proxmox VE. Packages for the command-line client are already provided for hosts running Debian or Debian…
featureAllow to read passwords from credentials passed down by systemd. Examples are the API token secret for the Proxmox Backup Server, or the password needed to unlock the encryption key.
featureImprovements to the vma-to-pbs tool, which allows importing Proxmox Virtual Machine Archives (VMA) into Proxmox Backup Server:
featureOptionally read the repository or passwords from environment variables, similarly to proxmox-backup-client.
featureAdd support for the --version command-line option.
securitySeveral vulnerabilities in GRUB that could be used to bypass SecureBoot were discovered and fixed (PSA-2025-00005-1) The Documentation for SecureBoot now includes instructions to prevent using vulnerable components for booting via a revocation policy.
+22 more items in the release notes3.3.7 Apr 5, 2025
fixfix #5982: garbage collection: add a check to ensure that the underlying file system supports and honors file access time (atime) updates. The check is performed once on datastore creation and on start of every garbage collection (GC) task, just to be sure. It can be disabled in the datastore…
othergarbage collection: support setting a custom access time cutoff, overriding the default of one day and five minutes.
otherui: expose flag for GC access time support safety check and the access time cutoff override in datastore tuning options.
otherdocs: describe rationale for new GC access time update check setting and the access time cutoff check in tuning options.
featureaccess control: add support to mark a specific authentication realm as default selected realm for the login user interface.
breakingfix #4382: api: access control: remove permissions of token on deletion.
fixfix #3887: api: access control: allow users to regenerate the secret of an API token without changing any existing ACLs.
fixfix #6072: sync jobs: support flags to limit sync to only encrypted and/or verified snapshots.
featureui: datastore tuning options: expose overriding GC cache capacity so that admins can either restrict the peak memory usage during GC or allow GC to use more memory to reduce file system IO even for huge (multiple TiB) referenced data in backup groups.
otherui: datastore tuning options: increase width and rework labels to provide a tiny bit more context about what these options are.
otherui: sync job: increase edit window width to 720px to make it less cramped.
otherui: sync job: small field label casing consistency fixes.
3.3.6 Apr 3, 2025
otherdatastore: ignore group locking errors when removing snapshots, they normally happen only due to old-locking, and the underlying snapshot is deleted in any case at this point, so it's no help to confuse the user.
featureapi: datastore: add error message on failed removal due to old locking and tell any admin what they can do to switch to the new locking.
featureui: only add delete parameter on token edit, not when creating tokens.
featurepbs-client: allow reading fingerprint from system credential.
otherdocs: client: add section about system credentials integration.
3.3.5 Apr 2, 2025
otherapi: config: use guard for unmounting on failed datastore creation
otherclient: align description for backup specification to docs, using archive-name and type over label and ext.
featureclient: read credentials from CREDENTIALS_DIRECTORY environment variable following the "System and Service Credentials" specification. This allows users to use native systemd capabilities for credential management if the proxmox-backup-client is used in systemd units or, e.g., through a wrapper…
fixfix #3935: datastore/api/backup: move datastore locking to '/run' to avoid that lock-files can block deleting backup groups or snapshots on the datastore and to decouple locking from the underlying datastore file-system.
fixapi: fix race when changing the owner of a backup-group.
breakingfix #3336: datastore: remove group if the last snapshot is removed to avoid confusing situations where the group directory still exists and blocks re-creating a group with another owner even though the empty group was not visible in the web UI.
featurenotifications: clean-up and add dedicated types for all templates as to allow declaring that interface stable in preparation for allowing overriding them in the future (not included in this release).
featuretape: introduce a tape backup job worker-thread option for restores. Depending on the underlying storage using more threads can dramatically improve the restore speed. Especially fast storage with low penalty for random access, like flash-storage (SSDs) can profit from using more worker threads.…
othergarbage collection: generate index file list via datastore iterators in a structured manner.
fixfix #5331: garbage collection: avoid multiple chunk atime updates by keeping track of the recently marked chunks in phase 1 of garbage to avoid multiple atime updates via relatively expensive utimensat (touch) calls. Use a LRU cache with size 32 MiB for tracking already processed chunks, this…
featurelogging helper: use new builder initializer – not functional change intended.
3.3.4 Mar 13, 2025
fixfix #6185: client/docs: explicitly mention archive name restrictions
otherdocs: using-the-installer: adapt to raised root password length requirement
otherdisks: wipe: replace dd with write_all_at for zeroing disk
fixfix #5946: disks: wipe: ensure GPT header backup is wiped
otherdocs: fix hash collision probability comparison
3.3.3 Feb 11, 2025
securityapi: datastore list: move checking if a datastore is mounted after we ensured that the user may actually access it. While this had no effect security wise, it could significantly increase the cost of this API endpoint in big setups with many datastores and many tenants that each have only access…
fixRevert "fix #5710: api: backup: stat known chunks on backup finish" due to a big performance impact relative to what this is protectign against. We will work out a more efficient fix for this issue in the future.
otherprune simulator: show backup entries that are kept also in the flat list of backups, not just in the calendar view.
otherdocs: improve the description for the garbage collection's cut-off time
otherpxar extract: correctly honor the overwrite flag
featureapi: datastore: add missing log context for prune to avoid a case where the worker state being unknown after it finished.
otherdocs: add synopsis and basic docs for prune job configuration
otherbackup verification: handle manifest update errors as non-fatal to avoid that the job fails, as we want to continue with verificating the rest to ensure we uncover as much potential problems as possible.
fixfix #4408: docs: add 'disaster recovery' section for tapes
fixfix #6069: prune simulator: correctly handle schedules that mix both, a range and a step size at once.
fixclient: pxar: fix a race condition where the backup upload stream can miss an error from the create archive function, because the error state is only set after the backup stream was already polled. This avoids a edge case where a file-based backup was incorrectly marked as having succeeded while…
3.3.2 Dec 10, 2024
fixfile-restore: fix regression with the new blockdev method used to pass disks of a backup to the isolated virtual machine.