Repository navigation
Releases
bgpcanary 1.7.7
Added
-
Transit lost behind an upstream. A new event for a network behind one of your upstreams that stops carrying you on from it while the upstream itself stays up, for example when your upstream loses its session with one of its own providers. The RIS peers that came in that way usually find another way in at once, so visibility holds and too few of them change upstream for an upstream-shift alert. Until now nothing was recorded. The event says:
- how many RIS peers that network usually carries you on to, and how many now;
- where those peers went, and whether visibility held;
- whether that network still carries you another way, for instance behind another of your upstreams. Then it may simply prefer that way in now.
It starts as informational and turns into a warning once it lasts (5 minutes by default), so a minute-long flap is recorded without alerting anyone. While it's informational it shows under Recent events on the dashboard, as a blue shade on the strip under the tree, and on the Events page with informational events shown. It ends once that network has carried at least half of its usual peers again for 10 minutes, so one that flaps is one event. One still gone after a day is closed as the new normal.
An upstream that loses almost all of its paths is still an upstream-down event, which names the networks behind it that stopped carrying you. Its alert carries the tree picture like the other routing alerts, and says so when RIPE's data looked unreliable at the time.
-
The transit-lost event is on by default and has its own settings, under Settings › Alert channels › What to alert on and under Detection & data sources:
- a tick box switches it off, for a network with so many upstreams or peers that it would be noisy;
- how big a share of a prefix's RIS peers a network must usually carry to be watched (10% by default);
- how far its peers must drop to count (to a quarter of its usual peers by default);
- how long it must last to become a warning (0 makes it a warning at once).
Setting its Notify choice to Off keeps the events and stops their alerts.
Fixed
- Settings › Detection & data sources: the line under Warn when this share of peers moves between upstreams read "undefined 83 of ~330 peers". It now says what the share means for each prefix: "a shift when 83 of ~330 peers move".
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.7 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.6
Fixed
- Event page: the list of raw BGP updates stopped at the event's end. It covered the five minutes before the start up to the end, so a short event's list left out the route changes that came right after it: a 15-second ping alert's list was empty although RIS heard 174 updates in the next five minutes. An instant event such as a new upstream chain showed only the five minutes before the change. The list now covers the same half hour either side as the rest of the page, and for a long event the half hours around its start and around its end, each with a link to what changed in it.
- Bookkeeping rows in that list wait behind a tick box that says how many there are. Looking-glass corrections, seed rows after a restart and stale withdrawals only correct the stored state, and on some events they made up most of the list.
- When the list reaches its limit it says so: how many updates the stretch holds, and a link to BGP data for the stretch. It used to stop at 3,000 rows with a "+".
- An event stored with its end before its start (releases before 1.3.3 could close one that way) now gets its page, its tree and Around this event from its start, instead of a window that stopped half a minute before it.
- Event pages find a stretch's stored updates by time, instead of going through the prefix's whole history first: an eight-day event's page came a second quicker.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.6 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.5
Added
-
Pins on an event's timeline. The strip under the path tree on an event's page now carries a pin for the event and one for each thing Around this event lists, so you can see when each of them happened and go there:
- blue where a burst of route changes began, with a faint band for as long as it lasted;
- red or orange for a RIPE Atlas ping round that stood out, and for another critical or warning event;
- grey for an info event (a new upstream chain, say) and for a stretch that wasn't heard live.
Point at a pin to read what it is. Click it, or drag the slider onto it, and a line under the slider says it in full for as long as the slider stays there. Pins that would sit on each other stack.
-
The line under the slider names shaded events too, on the dashboard's tree and on events replayed together: drag onto a shaded stretch or a mark and it says which event that is.
Changed
- An event's tree replays the same half hour either side as Around this event. A short event's replay covered ten minutes either side, so some of what the panel listed fell outside it.
- The strip's tooltip says what each entry is: an event, route changes, or a ping round.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.5 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.4
Added
-
Around this event. Every event page now opens with what else went on in the half hour either side of the event, whether or not it raised an event of its own:
- what the routes did: how many RIS peers changed their path, when, and where they went ("56 moved from AS64510 → AS64496 to AS64512 → AS64496"), and the peers per upstream at both ends of the window. A transit dropping out behind one of your upstreams shows here even when too few peers changed upstream for an upstream-shift alert;
- the RIPE Atlas ping rounds that stood out, each held against what's usual for that address (some probes never answer; that's allowed for), with a round whose replies took twice as long as usual marked too;
- every other event of that window, whatever its kind, the info ones and the ones found during a backfill included;
- the stretches bgpcanary didn't hear live, so a quiet line isn't taken for a quiet network.
A long event gets the half hours around its start and around its end. The same lines open the text report, so they reach the API's and the MCP server's report too.
-
A tree on every event page. An event that names none of your prefixes (a reachability loss on a carrier's link address, a Path finder watch) now shows the path tree replay, the charts, the "who carried" table and the raw updates of your prefix (with several, the one whose routes changed most around the event). That's usually where the cause of such an event shows: until now its page had no BGP on it at all.
Changed
- The strip under the tree shades every warning and critical event, not only outages, and the ones that name no prefix on every prefix's tree. Events that overlap share one shade. An event too short to see as a band (seconds, on a day's strip) gets a mark on top of the strip, and pointing at the strip says which events are there.
- Event pages load faster. The routes at the moments a page shows are read from the stored history once, where they were read once per moment: on a history of 240,000 updates an ordinary event's page took about half as long.
- The event page's "Related events" line is now the "Other events" part of Around this event.
Fixed
- Event page: the AS paths listed for the moment the upstream mix changed most were half a minute old. They were those of that bucket's start, while the numbers in the table above them are those of its end, so a sharp change showed the old paths under the new counts.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.4 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.3
Fixed
- The same past event listed several times, each marked "(backfill)". When RIS Live drops, bgpcanary notes the gap and fills it from RIPE's archive later. The gap was taken to begin at the last message's own timestamp, and RIS Live now and then passes on updates stamped days or weeks ago: when one of those was the last message before a drop, the "gap" reached back all those weeks, they were replayed, and every event in them was recorded again, once for each such drop. A gap now begins at the newest update the connection delivered (at most 15 minutes before its last message arrived), and a gap fill no longer records an event the Events list already holds, whatever stretch it replays. Repeats recorded by earlier versions stay in the list.
- "Last message" is when RIS Live's last message arrived, on the dashboard, in the data sources panel and in the API's
last_message, as documented. It was the message's own timestamp, so it could briefly read weeks ago; for the same reason the data sources panel could report RIS Live as "disconnected for" far longer than it had been. - A stretch that RIS Live missed could stay unfilled for good. bgpcanary lists the gaps RIS Live's drops leave and fills them from RIPE's archive every half hour. When RIS Live dropped again while the filler was at work (on a bad day it drops every few minutes), the new gap could be lost from the list: its stretch of the chart kept the line drawn while the feed was down, and nothing that happened in it reached the Events list. A gap was also crossed off when RIPE's archive failed to answer for it. A gap now leaves the list only once it has been filled, and as far as it has been filled; what's left is tried again half an hour later. A fill interrupted by a restart no longer replays the gaps it had already finished, which stored their updates a second time.
- An outage that wasn't, when RIPE's archive answered for some of your prefixes and not for others. bgpcanary asks RIPE's archive about each prefix by itself, the control prefixes included. When the request for one kept failing while the others went through, what did arrive was taken for the whole answer, and the prefix that was missing looked like one no RIS peer has a route to. The daily check against RIS's snapshot then withdrew every route it had: an outage alert for a prefix that was fine, until the looking glass put the routes back a few minutes later. A gap fill drew zeroes on its chart and recorded a "(backfill)" outage. A restart at such a moment recorded every route in the prefix's stored history as withdrawn long ago, so the tree and the event replays of the days before showed it without them (with the default control prefixes, that could happen to an install that monitors a single prefix). An answer is now used only for the prefixes it's for: the daily check and a restart leave the others as they are, and a gap fill waits until the archive answers for all of them.
- A week of history that RIPE's archive couldn't deliver was skipped without a word. A history load (a first start, a prefix you add later) or a long gap fill that got no answer for a week carried on with the next, on the routes as they stood before the missing week, or on none at all when it was the first: the charts and events of the weeks after it came out wrong, and nothing ever came back for the missing one. The load now stops there. A first start goes live with the history it has, and the rest is noted as a gap and loaded in the background once the archive answers, usually a few hours later.
- A route shown as withdrawn days before it was, in the tree and the replays of past moments. Once a day bgpcanary checks the routes it holds against RIS's snapshot and records a withdrawal for a route that went without one, dated right after the route's last update. The snapshot is about three hours old, and the check took every route not held at that moment for such a route: one withdrawn in those three hours, with a proper withdrawal, got a second one dated back to its last announcement, days or weeks earlier. The same happened to a route whose withdrawal RIS Live had missed and a gap fill had since recorded, and to one that went quietly after a restart had found it still there: the date used was the last update bgpcanary had heard itself, which can be older than what the history holds. After a start that got no snapshot for a prefix, the routes the looking glass hadn't brought back were recorded as withdrawn although RIS still had them. The check now leaves alone what was heard after the snapshot's time, asks the snapshot about a route bgpcanary doesn't hold, and dates a withdrawal it records after the newest update the history has for the route. Withdrawals dated back by earlier versions stay in the history.
- Routes withdrawn while bgpcanary was stopped were shown as withdrawn long before the stop. A restart checks the stored history against RIS's snapshot, which is about three hours old. After a longer stop the history ends before the snapshot's time, and every route withdrawn in between was recorded as withdrawn right after its last update, days or weeks earlier; the gap fill then stored the real withdrawal behind it. The daily check could do the same while a stretch RIS Live had missed was still waiting to be filled from RIPE's archive, and so could a first start whose history load had stopped early. A route that may have been withdrawn in a stretch still to be loaded is now left alone until it is loaded, and the next daily check records what is still unexplained then. On a day when RIS Live drops often, that can put off the history's clean-up of quietly dropped routes by a day; the routes bgpcanary counts and alerts on are checked as before.
- Three stretches that bgpcanary didn't hear were never filled in. What RIS Live misses is noted as a gap and filled from RIPE's archive a few hours later. Three kinds of stretch never made the list, so their part of the chart stayed as it was drawn while nothing was heard, and nothing that happened in them reached the Events list: the time between a RIS Live drop and a stop of bgpcanary that came before RIS Live was back; the time between a start and RIS Live's first sync, which on a new install includes the whole history load and otherwise lasts as long as RIS Live can't be reached; and, when RIS Live dropped again within seconds of reconnecting, the time from there to the next reconnect. All three are noted now and filled like any other gap.
Changed
- Container image: the base image (
python:3.13-alpine3.24) refreshed to its current build for the latest Alpine and Python fixes.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.3 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.2
Added
- Alpine Linux. The installer now takes Alpine 3.22 or newer alongside Rocky Linux, AlmaLinux and RHEL 9 and 10: the service runs under OpenRC's supervise-daemon as the unprivileged
bgpcanaryuser, nginx gets the same site in/etc/nginx/http.d, the certificate lives under/etc/ssl, and the log goes to/var/log/bgpcanary/bgpcanary.log, rotated weekly.bgpcanary-updateandbgpcanary-uninstallknow both init systems, and CI installs, updates and uninstalls on Alpine on every push. Thanks to tcroghan-LCT, whose pull request (#9) worked out what Alpine needs and tested it on fresh installs.
Fixed
- Outbound TLS trust: the bundled certifi is updated to 2026.10.23, which drops the emSign root whose intermediate CAs had compromised keys (GHSA-6q48-6c7w-9gw2). It covers the HTTPS calls to RIPEstat, RIPE Atlas, PeeringDB, GitHub and webhooks; RIS Live and email use the system's own trust store. (Proposed in #11.)
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.2 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.1
Added
- A built-in MCP server (
/mcp), so an AI assistant (Claude Code, Cursor, VS Code, Claude Desktop through a bridge, any MCP client that sends a header) can read your bgpcanary and answer "is anything wrong?", "what happened last night?" or "what is this network to me?". Off until an admin turns it on under Settings › MCP server; every user connects with their own API token, and it's read-only: the status, the events, one event with its report, a neighbouring network, a connection check. No new dependencies: it speaks the protocol's HTTP transport itself. Setup, the tools and troubleshooting: docs/MCP.md.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.1 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.7.0
Added
- A JSON API for your own scripts and dashboards (
/api/v1/), read-only, with per-user tokens: the monitor's status with each prefix's visibility, state and neighbours and how many events are open; the events with the Events page's filters, a prefix or a time window; one event with the alerts sent for it and its text report; and a neighbouring network with what it is to you and the events about it. Tokens are created under Settings › API tokens (shown once; only a hash is kept), carry their owner's role, and can be revoked there or, by an admin, on the Users page. The shapes are a promise: within v1 a field is only ever added. Documented in docs/API.md. - Tests (
tests/): the first automated tests, run by CI on every push. They hold the API's contract.
Fixed
- The tab icon and title follow the outage state again. Since 1.3.1 the "not live yet" banner's poll had taken over the
/api/statusaddress that the once-a-minute check used, so a tab left open kept the plain icon and title when an outage began (and the alert ones after it ended) until the page was reloaded. The banner polls/api/loadingnow, and/api/statusanswers with the monitor's status and each prefix's summary, as before.
Changed
- README: screenshots of the Path finder (the tree and the traceroute), BGP data (Paths and Changes), the Events page with its filters, an upstream-down event, several events replayed together and a neighbouring network's page; the connectivity tree screenshot now shows the "moved" and "reaches you another way" states.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.7.0 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.6.3
Added
- A page for each neighbouring network (
/neighbours/<ASN>): what it is to you with the one-click answer (Add as a peer, or one of your transits), how many RIS peers reach you through it right now, and the events about it. The dashboard's "unexpected" badge leads there, so filing a network as a peer no longer needs an open event.
Changed
- "What is this network to you?" now explains why a network you don't peer with directly can appear as your neighbour (route servers don't add their own ASN to the path, and some exchanges pass route-server routes on to partner exchanges), is shown to viewers too (the buttons stay admin-only), and its peer button reads Add as a peer.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.6.3 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.
bgpcanary 1.6.2
Added
- Events page: open or closed. A new first filter shows only the events still going on (no end yet) or only the ones that have ended, alone or together with the type and severity filters. The filters apply as soon as you change one; the count says what you're looking at ("3 open events"). Coming back to the page shows the filters its address says, not the ones the browser remembers picking. The dashboard's recent-events panel links to the open ones.
- Events page: a type in the list is clickable and filters to that type, keeping the other filters.
Changed
- Container image: the base image (
python:3.13-alpine3.24) refreshed to its current build for the latest Alpine and Python fixes.
Update a server install with sudo bgpcanary-update. Container image: ghcr.io/gregsowell/bgpcanary:1.6.2 (with Docker Compose: docker compose pull && docker compose up -d).
New installs: docs/INSTALL.md for a server, docs/DOCKER.md for a container.