* fix(postgres): make node traffic sync robust after public API inbound updates
The background NodeTrafficSyncJob (every 5s) started failing after a
successful POST /panel/api/inbounds/update/{id} (including flows that
inject streamSettings.externalProxy) with:
node traffic sync: merge for <node> failed:
ERROR: CASE types boolean and integer cannot be matched (SQLSTATE 42804)
Root cause:
- The merge lives in setRemoteTrafficLocked (called from SetRemoteTraffic).
- The client_traffics delta path used a dialect-sensitive expression:
enable = enable AND ?
last_online = GREATEST(last_online, ?)
- On PostgreSQL, GREATEST / AND / COALESCE are implemented with internal
CASE expressions. When "enable" columns (client_traffics, inbounds, ...)
were INTEGER (common after SQLite → PG data migrations, older
AutoMigrate, or mixed write paths) and the right-hand side was a
boolean parameter (from snapshot ClientStats or form-bound API payload),
PG rejected the expression at plan time.
- The public API update path (unlike the internal remote wire path)
always runs updateClientTraffics + UpdateClientStat + SyncInbound.
This touches client_traffics.enable rows for any inbound that has
clients.
- SQLite tolerated 0/1 numeric bools; PG is strict.
Fix:
- Use an explicit CASE with ::boolean casts in the critical enable
expression so the result type is always boolean.
- Make GreatestExpr emit safe casts on Postgres.
- Add a one-time normalization step in MigrationRequirements (runs on
startup + xray restarts) that forces the relevant enable/enabled
columns to boolean on Postgres using an idempotent DO block + USING
cast. This cleans up pre-existing skew without a full re-migration.
This branch is based on upstream/main (original mhsanaei/3x-ui main).
The node traffic sync now survives arbitrary public-API inbound
updates on PostgreSQL.
* fix: make client traffic enable merge expression safe on SQLite too
The previous commit introduced an explicit CASE for the "only node
can disable" logic in the node traffic sync merge to fix the PG
"CASE types boolean and integer cannot be matched" error after
public API inbound updates.
That expression used PostgreSQL-only `::boolean` casts:
CASE WHEN ?::boolean THEN enable::boolean ELSE false END
This is invalid syntax on SQLite (and would break the merge when
the client_traffics delta UPDATE runs — which is commonly triggered
right after an API /inbounds/update because that path calls
updateClientTraffics + SyncInbound and touches client_traffics rows).
Extracted the expression to a new dialect-aware helper
`ClientTrafficEnableMergeExpr()` (following the same pattern as
GreatestExpr, JSONClientsFromInbound, etc.).
- On Postgres: keeps the strict boolean-typed CASE with casts.
- On SQLite: uses a numeric-compatible form
`CASE WHEN ? THEN enable ELSE 0 END` that produces the expected
0/1 result matching the column affinity.
The logical behavior ("node may only force-disable, never re-enable")
is preserved on both databases.
This is a follow-up commit on the same branch so that one PR
contains both the original Postgres fix and the SQLite compatibility
fix.
Builds directly on top of 91643f68.
* fix
---------
Co-authored-by: Rqzbeh <rqzbeh@users.noreply.github.com>
Co-authored-by: Sanaei <ho3ein.sanaei@gmail.com>
* feat(mtproto): add MTProto (FakeTLS) protocol via managed mtg sidecar
Xray-core has no mtproto proxy, so mtproto inbounds run as standalone
mtg (9seconds/mtg) sidecar processes managed by the panel — one per
inbound — and are excluded from the generated Xray config entirely.
- model: MTProto protocol constant, validator, and FakeTLS secret
helpers (GenerateFakeTLSSecret/HealMtprotoSecret)
- mtproto package: per-inbound mtg process manager with reconcile,
graceful stop, and best-effort Prometheus traffic scraping
- runtime: delegate mtproto inbounds to the mtg manager instead of the
Xray gRPC API; skip mtproto when building the Xray config
- web: boot reconcile + StopAll wiring, periodic reconcile/traffic job,
port-conflict transport, secret healing on inbound add/update
- sub: tg:// proxy share-link generation
- frontend: protocol option, Zod schema, Protocol tab (FakeTLS domain +
regenerable secret), info-modal link, and i18n
- provisioning: fetch mtg v2.2.8 in install.sh, DockerInit.sh, and the
Linux + Windows release workflows
* fix
* fix
* fix: address Copilot review comments on mtproto PR
- web/web.go: create NewMtprotoJob once and reuse for cron + initial run
- mtproto/manager.go: StopAll cleans up per-inbound config files on shutdown
- mtproto/manager.go: CollectTraffic releases mutex before HTTP scrapes to
avoid blocking Ensure/Reconcile/Remove during network I/O
- database/model/model.go: panic on crypto/rand failure in mtprotoRandomMiddle
instead of silently producing a weak all-zero secret
- install.sh: fix chmod to handle renamed bin/mtg-linux-arm on armv5/v6/v7
Stop hand-writing OpenAPI response examples, which kept drifting from the real payloads (clients/traffic missing fields, inbounds/list exposing userId which is json:"-", the fictional inbound-443 tag instead of the real in-<port>-<transport> form).
tools/openapigen now emits frontend/src/generated/examples.ts: a per-struct example instance built from type defaults, validate oneof/min bounds, and example: struct tags, with nested-ref expansion and a cycle guard. build-openapi.mjs composes the {success,obj} envelope from it for any endpoint annotated with responseSchema (+ responseSchemaArray for lists); the hand-written response is dropped for those. Service DTOs InboundOption/ApiTokenView/ProbeResultUI are added to the walker.
#4996: client password regeneration now produces a valid Shadowsocks 2022 PSK (correct base64 length per cipher) when an SS2022 inbound is attached, in both the single and bulk client forms; backend surfaces ssMethod on /inbounds/options so the UI can pick the right length.
Also: Swagger UI persists the Authorization token across reloads (persistAuthorization).
* feat(nodes): add stable panel GUID identity (multi-hop phase 0)
Per-panel autoincrement node ids are meaningless one hop away, so in a chained topology (Node1 -> Node2 -> Node3) the master cannot attribute online clients or inbounds to the physical node that hosts them (#4983).
Introduce a stable self-identifier: each panel generates and persists a panelGuid (settings table, mirroring GetSecret), returns it in panel/api/server/status, and the master learns it per node via the heartbeat into a new Node.Guid column. Guarded so an old-build node or a failed probe never clears a known GUID. No behavior change yet - this is the identity foundation Phases 1-2 key on.
Refs #4983
* feat(nodes): attribute inbounds to their origin node by GUID (multi-hop phase 1)
Add Inbound.OriginNodeGuid: the GUID of the panel that physically hosts an inbound. Empty means this panel's own xray; set means it was synced from a node. SetRemoteTraffic now fills it per synced inbound - keeping a non-empty value the node forwarded from its own sub-node (so a transitive inbound stays attributed to the deepest node across hops), and otherwise attributing the node's own local inbounds to that node's GUID. Empty (old-build node without a GUID) leaves the existing node_id-based attribution untouched.
The field rides the existing inbound JSON, so /list propagates it up the chain with no serve-side change. Phase 2 will key per-node online off this instead of the panel-local node_id.
Refs #4983
* feat(nodes): key online status by node GUID end-to-end (multi-hop phase 2)
Replace the panel-local node-id keying of per-node online status with the stable panelGuid, so a client several hops down a node chain is attributed to the node that physically hosts it instead of the intermediate node it syncs through (#4983).
xray/process.go stores each direct node's reported GUID-keyed subtree and merges them (correct at any depth); the service assembles GetOnlineClientsByGuid (own clients under this panel's GUID + every node under its GUID). FetchTrafficSnapshot fetches the new /clients/onlinesByGuid, falling back to the flat /onlines for old-build nodes (keyed under the node's GUID or a master-local synthetic id). The node rollup, the WS onlineByGuid/activeInbounds fields, and the inbounds-page rollup all scope by GUID; local inbounds get their OriginNodeGuid filled with the panel's GUID at serve time so the frontend keys uniformly.
Old-build nodes degrade to the prior flat behaviour via the synthetic node:<id> key. Refs #4983
Refs #4983
* feat(nodes): surface transitive sub-nodes on the master (multi-hop phase 3a)
Each panel publishes read-only summaries of the nodes it manages via GET /panel/api/server/descendants (node API token). The heartbeat job caches each direct node's summaries; GetNodeTree merges them as transitive model.Node projections (Id 0, Transitive=true, ParentGuid = their parent node's GUID) and recomputes InboundCount/OnlineCount/DepletedCount per origin GUID so a direct node shows only its own inbounds and each sub-node shows its own (#4983).
The Nodes-page list endpoint and the heartbeat broadcast now return the tree; GetAll stays direct-only for probing/syncing. One transitive level is surfaced (covers Node1->Node2->Node3); deeper recursion is a follow-up. Backend only - the Nodes-page nested UI lands next.
Refs #4983
* feat(nodes): render transitive sub-nodes nested + read-only on the Nodes page (multi-hop phase 3b)
The Nodes page now shows a node's downstream sub-nodes (learned via the descendants tree) as indented, read-only rows ordered right under their parent: no enable toggle, probe, edit, delete, update, selection, or history expander - just a 'Sub-node' tag whose tooltip names the parent it is reached through. Desktop table and mobile cards both handle it. Transitive rows are keyed by GUID (their Id is 0) so they don't collide with real nodes (#4983).
Rows nest by parentGuid rather than AntD tree-children to avoid clashing with the existing per-row history expander. New labels added to en-US (other locales fall back until translated). Refs #4983
Refs #4983
* i18n(nodes): translate subNode/subNodeTip across all locales
Phase 3b added these two Nodes-page keys (read-only sub-node tag + tooltip) only to en-US; fill in the other 12 locales so the multi-hop sub-node UI is fully localized. The {parent} placeholder is preserved in every translation.
Refs #4983
The IP-limit job scrapes the Xray access log, which keeps lines tagged with a client's old email for up to a log-rotation cycle after a rename or delete. For each such email getInboundByEmail (settings LIKE %email%) found nothing, so the job logged 'failed to fetch inbound settings: record not found' every run and recreated an inbound_client_ips row for the dead email (rows reappeared even after manual deletion).
processLogFile now resolves the inbound once per email: if it maps to no inbound (gorm.ErrRecordNotFound) it logs at Debug, drops any orphan tracking row, and skips - so stale entries self-heal instead of spamming ERROR. The resolved inbound is passed into updateInboundClientIps, removing its internal lookup. updateClientTraffics also calls DelClientIPs alongside DelClientStat so a full inbound edit that drops an email doesn't leave a ghost row.
Closes#4963
The per-client traffic merge built enable = CASE WHEN ? = 0 THEN 0 ELSE enable END, mixing an integer literal with the boolean enable column. PostgreSQL rejects this with SQLSTATE 42804, aborting every node traffic merge transaction every 5s and freezing all up/down/last_online accounting on Postgres main panels. Replace with enable AND ?, which is type-safe on Postgres (boolean AND boolean) and identical in semantics on SQLite: the node may only disable a client, never re-enable one the panel already disabled.
Closes#4964
Node-backed client and inbound edits no longer hard-fail when the backing node is offline or disabled. Edits commit to the panel DB immediately and reconcile to the node when it reconnects (eventual consistency); the panel is the single source of truth for desired config.
- Add Node.ConfigDirty/ConfigDirtyAt; mark a node dirty when an edit commits without reaching it (cleared via CAS on ConfigDirtyAt after a full reconcile).
- nodePushPlan() reads node state fresh from the DB, skips the push for offline/disabled nodes (no 10s hang), and treats push failures as non-fatal across every mutation path (client add/update/del + bulk + attach/detach; inbound add/update/del/toggle/resetTraffic).
- ReconcileNode() pushes the panel's desired config to a node on reconnect (refreshing the remote tag cache first) and prunes node-side orphans; runs before the traffic pull in the node sync job.
- While a node is dirty the traffic pull applies only up/down deltas and node-initiated disables, never overwriting desired config from a stale node snapshot.
- Surface a non-blocking 'saved; will sync on reconnect' warning to the UI.
Validated with a two-panel Docker E2E: client delete/update, attach/detach, and inbound add/delete all reconcile correctly offline -> reconnect.
client_traffics is keyed by email (one shared row per client across every
inbound it is attached to). addClientTraffic filtered with
`inbound_id NOT IN (node inbounds)`, so when a client was attached to both a
node inbound and the mother inbound and the node inbound was attached first,
the shared row carried the node inbound's id (AddClientStat uses OnConflict
DoNothing and never refreshes it) and the local xray's traffic for that client
was dropped entirely. The client showed online but its usage stayed at zero
unless the mother inbound happened to be attached first.
Match purely by email instead. The reported emails come only from the local
xray, which only knows local-attached clients, so the query is still correctly
scoped, and this also repairs already-broken rows that a per-row AddClientStat
fix alone could not.
* fix(node-traffic): restart remote xray after disabling clients to kill active sessions
When a client's traffic limit is reached on a remote node, the panel pushes
enable=false to that node via UpdateInbound. The node calls RemoveUser on its
local xray, which blocks new connections but leaves any already-established TCP
session alive. The user could continue browsing/downloading until they
disconnected voluntarily.
Fix: after successfully pushing a client disable to a remote node, call
RestartXray on that node. This mirrors what already happens for the local node
when the "Restart Xray on client disable" setting is enabled (default: on),
and ensures active sessions are terminated immediately on all nodes where the
client was disabled.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor(node): restart remote xray after tx commit, not inside it
Move the remote RestartXray calls out of the addTraffic write
transaction. disableInvalidClients now returns the affected remote
node IDs instead of restarting their xray while the SQLite write lock
is held; AddTraffic performs the restart after the transaction commits
via restartRemoteNodesOnDisable. Avoids holding the serialized write
lock across slow per-node restart RPCs.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Sanaei <ho3ein.sanaei@gmail.com>
When a remote node syncs traffic back to the panel, the UPDATE in
setRemoteTrafficLocked wrote cs.Enable directly into client_traffics.enable.
If a snapshot carrying enable=true arrived after the central panel had already
set enable=false (due to the client reaching their traffic limit), it silently
re-enabled the client — letting them consume 2-3x their allotted quota before
the next disable cycle caught up.
Fix: replace the unconditional SET enable = ? with a CASE expression that only
allows a disable (0->0), never a re-enable (0->1). The central panel remains
the sole authority for turning a client back on.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Sanaei <ho3ein.sanaei@gmail.com>
* feat(sub): add finalmask support to JSON subscriptions
* feat(sub): modern xray JSON format with unified finalmask editor
Drop the legacy JSON subscription format entirely and always emit the
modern xray shape:
- Flatten proxy outbounds (no vnext/servers) for vless/vmess/trojan/
shadowsocks; hysteria was already flat.
- Express fragment/noise via streamSettings.finalmask instead of the
legacy direct_out freedom dialer + dialerProxy sockopt.
The global finalmask (tcp/udp masks + quicParams) is stored as a single
setting (subJsonFinalMask) and merged into every generated stream,
replacing the separate subJsonFragment/subJsonNoises/subJsonQuicParams
settings.
Reuse the existing FinalMaskForm (used by inbound/outbound) for the
settings UI via a small bridge component; add a showAll prop so all
TCP/UDP/QUIC sections render for the global case. This supersedes the
hand-rolled Fragment/Noises/quicParams tabs with the full mask editor
(all mask types).
Note: this is a breaking change — JSON subscriptions now require a
recent xray client on the consumer side.
* fix
---------
Co-authored-by: biohazardous-man <biohazardous-man@users.noreply.github.com>
Co-authored-by: MHSanaei <ho3ein.sanaei@gmail.com>
* feat(clash): add routing rules and enable routing option for Clash/Mihomo subscriptions
Allows adding custom YAML blocks and placeholders to Clash exports.
Why: Shifting routing to the client prevents server IP exposure for
DIRECT traffic and reduces unnecessary server bandwidth/CPU usage.
* fix
---------
Co-authored-by: Misfit-s <>
Telegram group chats can contain multiple bots. Commands addressed to another bot, such as /status@other_bot, should not be handled by the 3x-ui bot.
Closes#4893
Drive every client/inbound/group endpoint at 100k-200k clients on PostgreSQL and fix the latent issues found in previously-unbenchmarked paths:
- enrichClientStats: chunk the email IN lookup (was an unchunked bind that crashed past 65535 clients without traffic rows, taking down GetInbounds/GetInboundDetail/GetAllInbounds)
- GetOnlineClients: add the missing nil-process guard its siblings already have, so ListPaged no longer panics before xray starts
- GetClientTrafficByEmail: read UUID/subId from the indexed clients table instead of parsing the inbound's full settings JSON (439ms to ~1.5ms, flat in N)
- BulkResetTraffic: replace the per-email serialized loop with one chunked bulk UPDATE in a single transaction
- DelDepleted: delegate to the already-batched BulkDelete instead of deleting each depleted client one by one
Adds a postgres-gated full endpoint sweep plus an A/B benchmark, and SQLite correctness tests for the changed methods.
Bulk client operations bound their entire working set in a single
WHERE x IN (...) clause, which exceeds PostgreSQL's 65535-parameter limit
(and SQLite's 32766) and gives the planner a pathological query, so they
failed outright on inbounds/selections larger than the limit. Every such
query is now chunked at 400 items:
- BulkDelete / delete-all-clients: six IN queries chunked, and the
per-row delete tombstone (which swept the whole in-memory map on every
call, O(N^2)) replaced with a single bulk sweep.
- BulkAdjust: record and inbound-mapping lookups chunked.
- AddToGroup / RemoveFromGroup (bulk add/remove to group): three IN
queries chunked.
- replaceGroupValue (rename/delete group): inbound-mapping lookup chunked.
- List (all-clients listing): link and traffic lookups chunked.
Measured on PostgreSQL 16: delete-all-clients on a 100k-client inbound
now completes in ~7s (previously crashed at the parameter limit); bulk
add/remove to group ~6s and full client list ~1s at 100k.
sync_scale_postgres_test.go adds skip-gated benchmarks for delete-all,
group add/remove, and list.
Follow-up to the SyncInbound bulk rewrite, fixing the remaining O(M*N)
and O(M)-round-trip behaviour in the add/delete and bulk paths that made
them time out on large inbounds (worst case minutes), especially on
PostgreSQL.
- compactOrphans: chunk the "email IN (...)" lookup (400/batch) instead
of binding every email at once. A single huge IN exceeded PostgreSQL's
65535-parameter limit (and SQLite's) and made the planner pathological,
so add/delete failed outright past ~100k clients.
- emailsUsedByOtherInbounds: new batched form used by delInboundClients
(BulkDetach) and bulkDelInboundClients (BulkDelete), replacing a
per-email global JSON scan (O(M*N)) with one scan, and skipped entirely
when keepTraffic is set.
- BulkCreate: rewritten to validate/dedup in one pass, then group clients
by inbound and add them in a single addInboundClient call per inbound
(one getAllEmailSubIDs, one settings rewrite, one SyncInbound) instead
of running the full single-create pipeline per client.
- Bulk delete/adjust: batch DelClientStat/DelClientIPs with IN deletes
and wrap the settings Save + SyncInbound in one transaction, so the
per-row writes share a single fsync instead of one per row.
Measured on PostgreSQL 16 (one inbound, M=2000 affected clients):
- create: 8m35s (M=500) -> ~1-5s
- detach: 52s -> ~4s (flat in N)
- delete: ~16s -> ~1-4s
- adjust: ~20s -> ~7-10s
add/delete of a single client on a 200k-client inbound stays in seconds.
sync_scale_postgres_test.go adds skip-gated benchmarks (XUI_DB_TYPE=
postgres) for the single add/delete and the five bulk operations.
Every client mutation funnels through SyncInbound, which ran O(n) DB
round-trips per call: one SELECT per client, a Save+UpdateColumn per
client, and a per-row junction INSERT. Toggling a single client on a
large inbound issued thousands of queries and timed out, badly so on
PostgreSQL where each round-trip pays TCP latency.
SyncInbound now:
- loads existing records with a single chunked SELECT ... email IN (...)
instead of one query per client
- writes only the records that actually changed (skips no-op Saves), so
toggling/editing one client writes one row, not all of them
- batch-creates new records and batch-inserts the junction rows
Merge and sticky-field semantics are unchanged. Measured on PostgreSQL
16: a single-client toggle on a 50k-client inbound drops from ~8m54s to
~0.9s, and seeding 50k clients from ~2m48s to ~1.6s; 200k clients sync
in seconds.
A skip-gated benchmark (web/service/sync_scale_postgres_test.go, run
with XUI_DB_TYPE=postgres) reproduces and verifies the scaling.
UpdateUser and DeleteUser hit the node's email-based full-client endpoints, which fanned out to every inbound the client had on the node: editing a client wiped flow on the node's other inbounds, and detaching one node inbound deleted the client from all of them.
Make both inbound-scoped, mirroring AddClient. DeleteUser now detaches the resolved remote inbound id; UpdateUser passes an inboundIds scope so the node updates only that inbound.
Binary: extend the migrate-db subcommand with --dump and --restore so a
SQLite database can be exported to a portable SQL text dump and rebuilt from
one, alongside the existing --dsn PostgreSQL copy. Implemented in Go via the
bundled sqlite driver (new database/dump_sqlite.go); no external sqlite3 client
is required. Add ExportPostgresToSQLite (reverse of MigrateData) to build a
SQLite .db from live PostgreSQL data, reusing the shared copyAllModels helper.
Overview: add a "Download Migration" item to Backup & Restore plus a
getMigration endpoint/service that returns a .dump on SQLite or a .db on
PostgreSQL, so the data can seed a panel on the other backend. Document the
endpoint in api-docs and translate the three new strings across all locales.
Tests: cover the destination-side copy (AutoMigrate + copyTable into SQLite)
and the dump/restore round-trip including quoted values. Ignore *.dump.
The x-ui.sh helper that drives this from the CLI is in PR #4910.
Store API tokens as SHA-256 hashes instead of plaintext and return the token value only in the create response. List no longer exposes the token, and the UI drops the Show/Copy buttons in favor of a one-time reveal modal at creation.
Match hashes the presented bearer token before the constant-time compare, and a migration hashes any pre-existing plaintext rows in place so existing tokens keep authenticating. Docs and translations updated.
For an inbound deployed to a node, the button read the central panel's webCertFile/webKeyFile and inserted paths that don't exist on the node, crashing the node's Xray on startup.
Add a token-accessible GET /panel/api/server/getWebCertFiles that returns a panel's own web cert/key paths, Remote.GetWebCertFiles to fetch it from a node, and GET /panel/api/nodes/webCert/:id to proxy it. setCertFromPanel now calls the node endpoint for a node-assigned inbound and the local settings otherwise, warning instead of inserting wrong paths on error/empty.
Fixes#4854
Multi-inbound clients showed online on every inbound they were attached to. Xray's user-level traffic stat aggregates across all inbounds a client belongs to, so the email signal alone can't say which inbound was used.
Pair it with the inbound-level traffic signal under the same 20s grace and gate the per-inbound rollup on it: a client only shows online on inbounds that actually moved bytes this window. Remote nodes report no per-inbound activity and stay ungated (no regression). Adds GetActiveInboundsByNode, the activeInbounds WS field and POST /panel/api/clients/activeInbounds.
Fixes#4859
Custom geosite/geoip downloads built their own ssrfSafeTransport and never used the configured Panel Network Proxy, so geo updates failed on servers where GitHub is filtered. Route all custom-geo HTTP (startup probes + downloads) through panelProxy when set, falling back to the direct SSRF-guarded transport otherwise; the target URL stays SSRF-validated.
The Telegram bot only honored a socks5:// panel proxy and silently rejected http(s)://, despite the setting advertising both. Branch the fasthttp dialer (FasthttpHTTPDialer for http(s), FasthttpSocksDialer for socks5) and accept all three schemes in the fallback and NewBot validation.
Add tests proving the panel proxy is used by custom geo and that the bot dialer speaks HTTP CONNECT vs SOCKS5 per scheme.
client_traffics.inbound_id is a legacy single-inbound pointer that goes stale when an inbound is deleted and recreated: the email-keyed traffic row survives but references a missing inbound. Code that resolved the owning inbound from it broke several client operations.
- adjustTraffics: 'Start After First Use' (negative expiry) never converted to an absolute deadline on first traffic, so the countdown never started. Now resolves inbounds via the client_inbounds link and computes the new expiry once per email so multi-inbound clients stay consistent.
- GetClientInboundByEmail / GetClientInboundByTrafficID: fall back to client_inbounds when the pointer is dead, fixing reset traffic ('record not found'), client info, and Telegram set-tgId.
- autoRenewClients: resolve renew targets via client_inbounds so scheduled renews are not silently skipped.
- clients page: allow resetting a client with no inbound attachment (the backend already zeroes counters by email).
Add regression test for the delayed-start conversion under a stale inbound_id.
- Sample swap %, TCP/UDP connection counts and disk-usage % on the host ticker
- System History: Swap overlaid on the RAM tab, plus new Connections and Disk Usage tabs
- Persist the host time-series across restarts: gob snapshot beside the DB, written on a timer and at shutdown, restored on boot
- Live-refresh the open chart (2s for short ranges, 10s for longer)
- Localize CPU/RAM/Swap and the new tab/chart titles across all 13 languages and route legend series names through i18n
- Collect disk read/write and network packet-rate metrics on the host sampler
- Sparkline: optional 2nd/3rd overlaid series with a colored legend
- System History: merge Bandwidth (up/down), Disk I/O (read/write) and Load (1m/5m/15m) into single multi-line tabs
- Add a descriptive per-chart title and mobile-only tab icons to both modals
- Localize every chart title and tab label across all 13 languages
Relocate Remark Model & Separation Character from the General/Panel tab to the Subscription tab's Information section, beside Show Info and Email in Remark, since it only governs how share-link remarks are composed. The sample preview uses concrete example values and renders the separator literally.
Also drop the port from the subscription page link rows so each row shows just the inbound remark; the port still appears in the client QR modal and the client info modal.
Two multi-inbound client bugs from issue #4834:
- Clearing a client's reverse tag never persisted: SyncInbound keeps a non-empty sticky guard on reverse (shared with node-sync/rename), so the cleared value never reached the canonical clients.reverse column the edit form reads. Update now writes that column authoritatively from the submitted client, matching how it already writes email/updated_at directly.
- Attaching a new inbound reset xtls-rprx-vision: Attach seeded its wire client from the canonical clients.flow column, which a non-flow inbound can zero during the preceding update. It now derives the flow from EffectiveFlow (the per-inbound flow_override), so flow-capable targets keep the flow and others stay empty.
Adds service tests for both paths and a guard test confirming node-snapshot sync still preserves a stored reverse tag.
The panel's copy/QR share links are built client-side and fell back to window.location.hostname, so reaching the panel over an SSH tunnel (127.0.0.1/localhost) leaked localhost into the links - unlike the backend subscription path, which falls back to the configured Sub/Web Domain (issue #4829).
Expose webDomain/subDomain via /defaultSettings and add preferPublicHost: when the browser host is loopback, prefer the configured Sub Domain (then Web Domain) for share/QR links. An explicit node override or per-inbound listen still wins; a routable browser host is kept as-is.
Closes#4829
A setting row whose value column is empty or NULL (seen on some migrated databases) was parsed directly, so getInt/getBool and the GetAllSetting reflection path crashed with 'strconv.Atoi: parsing "": invalid syntax'. This made the Inbounds page (/defaultSettings -> GetPageSize) and the Settings page fail to load.
Treat an empty stored value the same as a missing row and fall back to the built-in default at the int/bool parse sites. String getters are unchanged, so legitimately-empty string settings stay empty.
Closes#4830
The inbounds page and Nodes page checked each client's email against a
single deduped union of every node's online clients, so a client connected
to one node showed as online on every inbound across every node. The local
online set was also derived from the email-keyed client_traffics.last_online
column, which remote-node syncs bump too, leaking remote-only clients onto
local inbounds.
Track online clients per node: the local panel's own xray clients under key
0 (derived from live traffic-poll deltas via RefreshLocalOnline, kept in
memory and independent of the shared last_online column) and each remote
node under its id. Add GetOnlineClientsByNode plus a /clients/onlinesByNode
endpoint and onlineByNode WS field; node.go and the inbounds rollup now scope
online by node. The flat GetOnlineClients union is kept for client-centric and
total-count views (Clients page, dashboard, telegram).
Closes#4809
xray-core >=26.5 makes the freedom finalRules context-aware: reverse-proxy traffic defaults to "block all targets". The template seeded finalRules with only allow geoip:private, so a bridge could not exit to WAN and reverse proxy silently broke
Switch the default direct freedom to a no-condition allow rule, the documented way to restore pre-policy behavior. Unlike an ip-based rule (0.0.0.0/0 or !geoip:private), it does not force per-connection OS DNS resolution under domainStrategy AsIs, so happyEyeballs/AsIs pass-through stay intact. LAN is still blocked by the geoip:private->blocked routing rule, and removing that rule still regains LAN access
Note: only affects new configs; existing installs keep their stored finalRules until reset or a follow-up migration.
SyncInbound runs once per inbound and unconditionally overwrites the canonical clients.Flow column. A non-flow inbound (Hysteria, WS, gRPC) strips flow to "", so when it syncs after a VLESS Reality inbound the column is wiped, and the hydrate endpoint returned that empty value — the edit form loaded a blank flow for multi-inbound clients (#4792).
Derive the hydrate flow from the first flow-capable client_inbounds.flow_override instead, which is always correct and order-independent. A non-empty guard in SyncInbound was rejected because it would make flow impossible to clear.
Closes#4792
The pinnedCertSha256 form field unmounts for non-pin TLS modes, so antd dropped it from the onFinish values and Zod rejected the missing string (the user-facing "invalid input"). Make it optional with a default so saving works in every TLS mode.
Saving now runs the connection test first and only persists when the probe is online; the add/update endpoints enforce the same probe so an unreachable node cannot be stored via the API either.
Selecting the http scheme forces TLS verify mode to skip and disables the control, normalized on open for existing http nodes.
http-vs-https probe failures report a clear "set the node scheme to http" message across the test button, save, and the backend gate.
Closes#4794
BulkDetach removed one client per (email x inbound) pair, each with its own
settings rewrite, transaction and full SyncInbound. Add delInboundClients to
remove all targeted clients from an inbound in a single pass and group removals
by inbound, turning O(emails x inbounds) write cycles into O(inbounds).
BulkAttach ran the global getAllEmailSubIDs scan once per target inbound via
checkEmailsExistForClients. Compute that snapshot once per call and thread it
through a new internal addInboundClient; the duplicate check is unaffected
because attach reuses each client's existing identity (same subId).
Covered by bulk_clients_test.go: VLESS round-trip (linkage, settings JSON,
idempotency, record survival), skip-unattached, and Trojan key matching.
FetchCertFingerprint must accept any certificate by design: it fetches a
not-yet-pinned node's leaf cert (trust-on-first-use) so the admin can pin
it. Disabling verification is inherent to that, so go/disabled-certificate-check
cannot be cleared by code changes. Suppress the finding inline, matching the
existing lgtm convention in custom_geo.go.
FetchCertFingerprint read the leaf certificate from a bare insecure TLS
handshake, which CodeQL flagged as go/disabled-certificate-check. The
function intentionally accepts any cert (trust-on-first-use, so the admin
can pin a not-yet-trusted node), so verification cannot be enabled.
Capture the leaf cert inside a VerifyConnection callback instead, matching
the existing pattern in nodeHTTPClientFor that already clears the same
query. Behavior is unchanged.
Adds a per-node TLS verification mode to the Add/Edit Node dialog so the panel can reach nodes that serve HTTPS with a self-signed certificate:
- verify (default): normal CA validation.
- skip: InsecureSkipVerify, with a clear UI warning that it drops MITM protection.
- pin: validates the leaf certificate's SHA-256 (base64 or hex) via VerifyConnection while bypassing the default chain/name check — keeps MITM protection for self-signed certs, the secure alternative to skip.
New Node model fields tlsVerifyMode + pinnedCertSha256 (gorm auto-migrated). Probe() selects the HTTP client per node via nodeHTTPClientFor, keeping the SSRF-guarded dialer. A new POST /panel/api/nodes/certFingerprint endpoint (FetchCertFingerprint) lets the UI fetch and pin the node's current certificate in one click. Endpoint documented in api-docs/openapi; i18n added across all locales. Verified end-to-end in Docker (verify rejects, skip bypasses, fetch matches, pin accepts correct / rejects wrong).
UDS listen already worked for proxying (the listen string is passed to xray verbatim and port 0 is accepted), and the Go sub/link layer already ignores the bind listen. The only gap was the frontend resolveAddr, which would put a socket-path listen into share/sub links (e.g. vless://uuid@/run/xray/x.sock:0). resolveAddr now treats a path-style listen (starting with / or @) as having no client-reachable address and falls back to hostOverride/hostname. Adds a test and a Listen-field help hint across all locales.
Since v3.1.0 every fallback row had to reference a panel inbound via childId, so rows with only a free-form dest (e.g. 8080 or 127.0.0.1:8080 to an external Nginx) were silently dropped at three layers: the frontend save filter, the backend SetByMaster guard, and BuildFallbacksJSON. A row is now valid when it has a child OR an explicit dest; self-references normalize to childId 0, and BuildFallbacksJSON prefers an explicit dest (also fixing rows whose child was deleted). UI gains allowClear on the child picker; help text updated across all locales. Verified end-to-end in Docker: a free-form dest fallback now persists and is injected into the live xray config. Refs #4554, #4639.
RegWarp now stores config.client_id from the Cloudflare registration, and WarpModal sources the reserved bytes from the live config response (falling back to stored creds). Previously reservedFor read an always-missing client_id, producing an empty reserved array.
A client shared across multiple nodes has a single email-keyed client_traffics row, but each node reports its cumulative up/down. setRemoteTrafficLocked overwrote the row with one node's cumulative, so non-owning nodes hit the create branch and OnConflict-DoNothing, silently dropping their traffic and under-counting the client.
Make the shared row a pure accumulator (like the local path): a new node_client_traffics(node_id, email) baseline table stores each node's last cumulative; the node path converts cumulative to a per-node delta (clamped to the post-reset value on a negative delta) and does up = up + delta. First observation seeds the baseline and adds 0 so upgrades and newly-shared clients are not double-counted. Create-vs-accumulate now keys off global email existence. Baselines are cleaned in DelClientStat, the node sweeps, and NodeService.Delete.
A non-empty, non-any Address (listen) leaked into the tag as
in-<listen>:<port>-<transport> (e.g. in-127.0.0.1:443-tcp). The tag is
now always in-<port>-<transport>, with the node prefix and numeric dedup
suffix still handling uniqueness across nodes and same-port/different-listen
inbounds. Mirrored in the Go authority and the TS form preview, kept in
parity by tests.
Existing colon-form tags are now treated as custom, so editing such an
inbound preserves its tag rather than rewriting it; new inbounds (or a
cleared tag field) get the clean form.
Reject creating or editing a client with a subId already owned by a different client, mirroring the email-uniqueness checks against client_records in Create and Update (BulkCreate inherits via Create). The old multi-inbound model duplicated a client across inbounds sharing one subId, so this check was dropped; the first-class multi-client model makes per-client subId uniqueness correct again. Existing duplicates are left untouched; only new/edited duplicates are blocked.
Adds the ability to update node panels to the latest release from the Nodes
page: select online, enabled nodes (checkboxes) and trigger their official
self-updater, or use the per-row Update action. A node whose reported panel
version trails the latest GitHub release is flagged with an 'update available'
tag (compared via lib/panel-version, mirroring the Go isNewerVersion).
Backend: Remote.UpdatePanel calls the node's existing
POST /panel/api/server/updatePanel; NodeService.UpdatePanels fans out over the
selected ids, skipping disabled/offline nodes with a per-node reason; exposed
as POST /panel/api/nodes/updatePanel (documented in endpoints.ts + openapi.json).
The bulk request sends a JSON body, so it sets Content-Type: application/json
explicitly — axios defaults POST to form-urlencoded, which made ShouldBindJSON
fail with 'invalid character i'.
Also reuses the clients-page online cue on the Nodes page: a pulsing green dot
plus green label for an online node. The .online-dot style moved to the shared
styles/utils.css so both pages load it.
Translations for all new node keys added across every language file.
The subscription page leaked an inbound's server-side Listen IP into the
client-facing URLs when a bind address was set:
- Per-config links: resolveInboundAddress returned the bind Listen IP
(loopback/private/public alike) instead of the host the subscriber
reached the panel on. It now returns the node address for node-managed
inbounds, otherwise the subscriber host; the bind Listen is ignored
(External Proxy remains the way to advertise a specific endpoint).
- Subscription Copy URL (SUB/JSON/CLASH): BuildURLs composed the base
differently from the panel's Client Information page and never
normalized the request host, so a loopback/bind request leaked the raw
IP. The composition is extracted into the shared
SettingService.BuildSubURIBase, used by both the panel and the sub page
so they render identically, and fed the already-normalized subscriber
host.
Auto-generated inbound tags (in-<port>-<l4>, n<id>- prefixed for node inbounds) now re-derive when port/listen/transport change on update instead of keeping the stale round-tripped value. The resolved tag is mirrored onto the API response, and NodeID is pinned to the stored row so a node inbound never loses its n<id>- prefix on edit. The edit form recomputes the tag live via a Go-parity helper so the JSON preview matches what gets saved.
Make node/central tag matching prefix-agnostic in all three places (traffic attribution, remote-id resolution, and the orphan sweep) so an n<id>- prefix present on only one side can no longer spawn duplicate inbounds or drop traffic on sync.
Force LF on shell scripts via .gitattributes (CRLF broke the Docker build shebang when the repo is checked out on Windows) and add a .dockerignore to keep node_modules/.git out of the build context.
Adds Go and frontend tests covering tag re-derivation, prefix-agnostic matching, and node-snapshot prefix mismatch.
UserLoginNotify ran SendMsgToTgbotAdmins synchronously on the login request goroutine. When Telegram was unreachable, the send retried up to 3x with a 30s timeout each, blocking the login handler for ~90s+ and effectively locking users out (issue #4585).
Dispatch the send in a goroutine after the cheap bot-running/login-notify-enabled guards so login always returns promptly; the existing per-send 30s context timeout and bounded retries keep the background goroutine from leaking.
When an inbound is deleted and recreated it gets a new id, but the shared-by-email client_traffics row keeps the old (now deleted) inbound_id because AddClientStat's OnConflict-DoNothing never refreshes it. The traffic updater matched rows with inbound_id IN (local inbounds), so those orphaned rows were dropped: client traffic and online status stopped updating and auto-renew skipped them, while inbound-level traffic (matched by tag) kept working and the client count still showed (matched by email).
Match by email and exclude only rows owned by a node inbound (inbound_id NOT IN (node inbounds)) in addClientTraffic and autoRenewClients. The local Xray only reports local-client emails, so a stale local pointer no longer hides the row, while genuine node-owned rows stay protected. Verified against a real affected dump: visible rows went from 4/668 to 668/668.
