Logs
Everything the PBX records about what happened is in one place: Monitoring and reports > Logs in the admin console's sidebar (/admin/logs opens the first log you may read). Each log is a tab of that page: Activity, one timeline over all of them, the audit logs, the SIP logs, the webhook deliveries and the System log, the recent warnings and errors of the server.
| Tab | What it shows | Who sees it |
|---|---|---|
| Activity | One searchable timeline: configuration changes, sign-ins, upgrades, every other audited action, webhook deliveries and alerts. | audit:read (company), system.audit:read (whole system) |
| Audit log | Every audited action with its actor, channel, address and outcome. | audit:read, system.audit:read |
| Audit log > Access audit | Sign-ins and refusals, API key use, recording downloads. | audit:read, system.audit:read |
| Audit log > Change history | Before and after of each configuration change, with revert. | audit:read |
| SIP trace | Capture the SIP messages of a trunk, extension or number for a short time, with a live ladder and pcap download. | diagnostics:read, diagnostics:write to start |
| SIP events | Every REGISTER, INVITE, MESSAGE and BYE with its final response. | audit:read (company), system.settings:read (traffic of no company) |
| Webhook deliveries | Outbound webhook deliveries and the dead letters, with redelivery. | integrations:read |
| System log | Recent warnings and errors of the PBX server. | system.settings:read |
The alerts are a tab of Monitoring and reports > Health and alerts (Alerts). How long the change history is kept (History retention) is under Settings > Storage, the log levels (Log levels) under Settings > System. The call history stays in the workspace sidebar (Calls); upgrades and rollbacks keep their page under Settings > System, and their requests also appear in Activity.
Activity
Activity merges the existing records into one list, newest first. Nothing is copied: each row is read from its own store when you open the page.
- Configuration change, Sign-in, Upgrade and Audit: rows of the audit log (
config.*,auth.*,upgrade.*and every other action). - Webhook delivery: the company's outbound webhook deliveries (delivered, waiting, retrying, dead), shown when you also hold
integrations:readfor the whole company, and only for the events you may see. - Alert: the alerts of the alert centre (whole system only, with
system.settings:read).
Every log opens with the same layout: the title with the number of rows loaded, a search box and filter chips under it (the choices are kept in the page address, so a filtered view can be bookmarked or shared; Clear filters resets them), then the table. The table loads 50 rows at a time (Show more adds the next page) and shows a No matches message when the filters leave nothing. On a phone each row is a card.
Filters:
- Time range: the last hour, 24 hours, 7 days or 30 days, or a custom range. One search covers at most 31 days.
- Source: one of the kinds above.
- Severity: Info (a success, a delivered webhook, an informational alert), Warning (a refused action, a webhook being retried, a warning alert) or Error (a failed action, a dead webhook, a critical alert).
- Actor: a user id. Rows without an actor (deliveries, alerts) are left out while it is set.
- Search: text in the action, the alert rule or summary, or the webhook event type and endpoint name.
The actor and the two dates of a custom range sit in a small row under the filter chips.
Each row has a Details link to the page that holds the full record: the change history for a configuration change, the access audit for a sign-in, the upgrades page, the audit log, the alerts or the webhook deliveries.
Export CSV downloads the rows of the current filters (at most 5000 rows, newest first). Cells that a spreadsheet could read as a formula start with an apostrophe.
A company administrator sees their own company. Someone who also holds a system role can switch Scope to Whole system: the records that belong to no company (system roles, upgrades and rollbacks, company creation) and the alerts. A company's rows never appear in the system view, and the reverse.
API: GET /api/v1/activity and GET /api/v1/system/activity (list, with the usual filter, limit and cursor), GET /api/v1/activity/export and GET /api/v1/system/activity/export (CSV).
Audit log
The Audit log tab lists every audited action, newest first: when, what (a readable label with the action code under it), who and through which channel (web, API, command line, AI client or the system itself), the object and the outcome. People and objects are shown by name: the user, API key or device that acted, and the extension, department, user or other object it acted on (an extension reads Front desk · 106). An object deleted since keeps the last name the change history recorded. When no name is known (a scheduled job, a failed sign-in for an unknown account, a whole-system record) the row shows the kind and the start of the id; the full ids are always in the row's details.
The search box finds text in the action (auth., login). The Outcome, Actor and Channel chips narrow the list; More filters holds the From and To days, the Object type and the Object id (on a phone it opens as a sheet from the bottom of the screen). The days are your own local days, from the start of the first to the end of the last. All of them are kept in the page address.
API: GET /api/v1/audit-log returns actor_name and entity_name next to the ids when they are known.
SIP events
SIP events lists every REGISTER (first registration, refresh, unregister), INVITE (new call and re-INVITE), MESSAGE and BYE that the PBX received or sent, newest first, one row per request with its final response:
- the time, the method and kind, the direction (in: reached the PBX, out: sent by it), the Call-ID and CSeq;
- From, To and the request address as
user@domain, the extension when the domain belongs to your company; - the other side's IP address, port and transport, and its software (User-Agent);
- the response code and reason (
486 Busy Here), or No answer when none came within 32 seconds; - the sign-in result: OK (accepted with credentials), Challenged (asked for credentials, 401/407), Failed (credentials refused, 403) or None;
- for a REGISTER the expiry it asked for; for a BYE how long the call lasted and the hang-up cause (
Reasonheader).
Rows are coloured by response: green for 2xx, grey for 1xx and 3xx, amber for 4xx, red for 5xx, 6xx and no answer.
Filters: time range (the last 15 minutes, hour, 6 hours or 24 hours, or a custom range of at most 24 hours), method, response class, extension, address (any part of From, To or the request address), IP address, Call-ID and text. Live reloads the newest rows every 5 seconds. Export CSV downloads the current filters' rows (at most 10000).
The list is a standard table with a filter bar. Open a row's ... menu and choose its first item to read the request and every response (and the challenge it answered) in a window, with links to the trace and the call.
In the call history, a call's details have a SIP events of this call link (for those who read the logs): it opens this page on the INVITE and BYE events of the call's legs.
Each row opens to show the request and every response with all their headers, and the PBX cause when the PBX refused a call: see SIP trace and reading SIP events. Message bodies (the text of a MESSAGE, SDP offers and answers) are not recorded unless a company turns on Keep message bodies. Passwords, digest responses, credentials in headers and media keys are always masked, and a password written in a SIP address (user:password@host) is dropped.
A company administrator sees their company's traffic. Someone with a system role can also switch to Traffic of no company: requests for unknown domains or the bare IP address, typically scanners.
Retention and volume
[sip_events]
enabled = true # needs ClickHouse
retention_days = 7 # 1 to 90
max_per_company_per_day = 2000000 # rows per company per UTC day
register_refresh_sample = 1 # keep 1 in N successful REGISTER refreshes
queue = 20000 # events waiting to be written
- Each row carries its own expiry (its time plus
retention_days), and ClickHouse deletes it after that; changing the setting applies to new rows. - Recording never slows calls down: the SIP edge only copies the header lines of the messages above into a bounded queue, and a background task matches requests with their responses and writes them to ClickHouse in batches. When the queue or the writer is full, events are dropped and counted (
mpbx_sip_events_dropped_total{reason}). - A company that sends more than
max_per_company_per_dayevents in a UTC day stops being recorded until the next day (counted ascompany_cap). - REGISTER refreshes are the bulk of SIP traffic (a phone refreshes every few minutes). They are all kept by default; with
register_refresh_sample = Nonly one in N successful refreshes is written (failed ones, first registrations and unregisters always are). Per-device aggregation was not chosen: a sampled row is still a real event with its own time and address, which is what you need when a phone's registrations start failing. - The page shows the SIP traffic of an all-in-one install (
mpbx runwith the SIP edge in the same process). With separate edge nodes, see the follow-up note FU-LOGS-3.
API: GET /api/v1/logs/sip and GET /api/v1/system/logs/sip (filter on ts, method, direction, extension, aor, ip, status_class, call_id, pbx_call_id, auth, text; company on the system route), GET /api/v1/logs/sip/export and /api/v1/system/logs/sip/export (CSV).
System log
The System log shows the last warnings and errors the PBX server logged, newest first: time, level, component (the module that logged it) and message. Filter by level, time window, component and text; More loads older events.
- The server keeps the last 2000 warnings and errors in memory, each message cut at 2 KiB. A restart empties the list; the full logs are still in the journal or the container logs (and in a support bundle).
- Messages are masked like the support bundle before they are shown: passwords, tokens, keys, credentials in addresses and telephone numbers never reach the page.
- Only what the active log level lets through is recorded. With the default level (
info) every warning and error is kept; to see more detail for a while, use Log levels. - The page shows the API server process. The media process and other nodes of a cluster keep their own logs, read them with
journalctlordocker logs.
API: GET /api/v1/system/logs (level, target, q, since, before_seq, limit), system scope with system.settings:read.