100 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
justme	49e787a7a5	backend: confirm a registered update by itself after a delay when auto update is switched on [CI]	2026-10-22 10:59:32 +08:00
justme	4e7a8c2ba3	web: draw isolated points as a connected series and mark unrecorded time in the adaptive audit [CI]	2026-10-22 10:57:53 +08:00
justme	93abf59afb	backend: leave the time before the first rate record unknown in the adaptive audit	2026-10-22 10:57:53 +08:00
justme	20b0f7fbd8	measurement: record the start of adaptive probing as a rate change	2026-10-22 10:57:53 +08:00
justme	af385cf688	web: show the adaptive probe-rate audit on the line page [CI]	2026-10-22 10:44:46 +08:00
justme	9ad34acbac	backend: store adaptive rate changes as events and serve an audit of them per line	2026-10-22 10:44:46 +08:00
justme	d63c3bff74	agent: keep adaptive rate changes apart from the logs and report them in the heartbeat	2026-10-22 10:44:46 +08:00
justme	557b169dae	web: make loss and retransmission charts obvious, draw sparse series as points and keep percent axes within 100% [CI]	2026-10-22 10:14:38 +08:00
justme	4a9608ead4	fleet: count only admitted peers against the session limit so silent connections cannot lock the roster out	2026-10-22 10:14:38 +08:00
justme	fee236fe56	web: explain the public-only default when an agent refuses a target [CI]	2026-10-22 07:36:29 +08:00
justme	32547032d8	agent: allow only public addresses as single-agent targets unless allowed_target_cidrs says otherwise	2026-10-22 07:36:29 +08:00
justme	8efc16f341	web: group the QUIC and TCP rows of a line card by protocol [CI]	2026-10-22 07:11:33 +08:00
justme	caee03b8cc	web: show single-agent targets on the visitor page and open a detail page from the card [CI]	2026-10-22 07:02:31 +08:00
justme	7ad226f29c	web: split the task page into agent-to-agent and single-agent tasks with templates and run status	2026-10-22 07:02:31 +08:00
justme	3a9ea46d6f	backend: serve single-agent task templates with their status and results, rename the kinds in storage	2026-10-22 07:02:28 +08:00
justme	e77880d372	measurement: name the kinds by protocol and action, add the http_request kind and QUIC settings	2026-10-22 07:02:23 +08:00
justme	bf2a598891	web: toast passkey registration failures	2026-10-22 05:59:56 +08:00
justme	758b6b1566	web: merge the per-agent lines into one table on the agent page	2026-10-22 05:59:56 +08:00
justme	f62eefd365	web: manage links in a master-detail explorer instead of the board	2026-10-22 05:59:56 +08:00
justme	70fc033789	tests: bring the slow mesh and fixed-rate cases in line with one line per pair and adaptive probing ... The mesh test expects the single line of a pair, dialed by the smaller id. Tests that send at a fixed rate pin adaptive to false, so the admission limit follows that rate.	2026-10-22 05:39:16 +08:00
justme	5822285a88	fleet: require a minimum version of the peer at the handshake [CI] ... MoShiro__MinAgentVersion makes the backend refuse agents below it at hello (fleet_version_required); fleet.min_server_version or MOSHIRO_MIN_SERVER_VERSION makes the agent drop a backend below it. Either side waits five seconds and shakes hands again, because the other one may be updating.	2026-10-22 05:28:29 +08:00
justme	fb3a9c260f	fleet: send log counters in heartbeats and pull the text on demand ... Heartbeats carry per-level counts, the most frequent log types (a stable hash of the message) and the newest id, no text. The node keeps its last 2048 lines; moshiro.logs.fetch lets the backend pull them. The log page pulls when it is open, and the backend pulls warnings and errors once a minute after a node reports an error.	2026-10-22 05:28:29 +08:00
justme	4c31a4a79e	agent: keep backing off on lines that stay down ... The control layer keeps doubling after the first minute up to ten minutes, and the QUIC redial doubles after twelve short attempts up to five minutes, instead of retrying every minute or every five seconds forever.	2026-10-22 05:28:29 +08:00
justme	e0376b5e33	measurement: adapt the echo probe rate to loss and RTT, and weight samples by the time they stand for ... udp_echo and tcp_echo run at the lowest rate that still proves the loss bound and keeps the RTT p99 confidence interval within tolerance, go to the maximum when the line loses packets or its RTT rises, and back off exponentially when it is unreachable. Every echo row carries a weight (the time it stands for, in grid intervals) and the analysis ops accumulate loss and RTT by it, so p50/p95/p99 and loss stay time-averaged when the rate changes. State changes are logged for audit.	2026-10-22 05:28:28 +08:00
justme	57acecdb86	web: show how long a line has been out of contact ... The Backend now keeps the time since a link's control layer was last up, because retries move the control state in and out of backoff and its own timestamp keeps resetting. The overview exposes it as control_down_since and the card shows it next to the lost-contact notice.	2026-10-21 17:16:09 +08:00
justme	5245f03a78	web: filter lines by agent with a searchable multi-select, and let the line cards adapt to their width [CI] ... A line shows when either end is selected; the selection is kept in ?agents=. Cards use container queries to tighten the columns and drop the hint text when narrow.	2026-10-21 16:59:13 +08:00
justme	08f261293c	web: equal-width cells, grey in-progress bucket and a time axis under each direction on the line cards ... Rows are laid out as protocol, cells and value on one line per metric. Each direction gets its own axis with ticks on the hour; cells without data and the bucket still being filled are grey.	2026-10-21 16:59:13 +08:00
justme	7fe9775f97	web: reset links to the default mode, and keep CN agents dial-only by default [CI] ... POST /api/admin/links/reset clears every manual link setting and turns automatic linking on. In the default rules a listening node in CN ranks below a listening node elsewhere, so the CN side dials; the links are pushed again when a node's country is first resolved or changes.	2026-10-21 16:35:58 +08:00
justme	35207a70ea	web: group the line card rows by direction so each agent pair is named once	2026-10-21 16:31:22 +08:00
justme	cfe6571db5	web: name the two agents on each direction row of the line cards	2026-10-21 16:29:56 +08:00
justme	9f5a6cf77a	analysis: split TCP retransmissions into forward and return, and show both on the line cards ... tcp.retrans.up counts retransmitted segments sent by the dialing side, tcp.retrans.down those sent by the listening side. The cards show RTT as a round trip (⇄) with forward (→) and return (←) rows for QUIC loss and TCP retransmission.	2026-10-21 16:26:34 +08:00
justme	48ffa56285	web: one line per node pair and RTT/loss on the link board ... A pair of nodes now has a single line, dialed by the non-listening side or by the smaller id when both listen, so RTT is measured once. The board draws it undirected and colours it by RTT and loss; a manual setting applies to the pair.	2026-10-21 16:20:38 +08:00
justme	e0f6667cd7	ci: build the backend image in its own workflow and shorten the panel offer wait [CI]	2026-10-21 15:49:38 +08:00
justme	5b1bf09eb0	ci: drop the backend e2e step from the dev image pipeline	2026-10-21 15:45:29 +08:00
justme	63f62956d0	ci: skip buildah unshare when the cache cleanup runs as root	2026-10-21 15:42:35 +08:00
justme	15318de919	Restore packaging/openwrt and the Agent updater paths swept into the previous commit [CI]	2026-10-21 15:37:57 +08:00
justme	fae05ff303	Rebuild dev images from a clean host [CI]	2026-10-21 15:36:32 +08:00
justme	099c34c715	Raise build parallelism for the 44-core CI host [CI]	2026-10-21 15:34:41 +08:00
justme	93ecc53cd0	verify-backend: do not leave reusable MSBuild nodes behind	2026-10-21 15:30:51 +08:00
justme	b56fd1a3cd	Allow four concurrent NativeAOT compiles [CI]	2026-10-21 15:27:04 +08:00
justme	c0f55275d6	verify-backend: report who holds the cache lock and keep build servers from inheriting it	2026-10-21 15:25:20 +08:00
justme	334b96324d	guard-step: stop the command when Woodpecker cancels a step with SIGKILL [CI]	2026-10-21 15:20:24 +08:00
justme	d1ece67a6a	Build the frontend once per pipeline and reuse it in the dev backend image [CI]	2026-10-21 15:17:37 +08:00
justme	3654d37da5	Limit concurrent NativeAOT compiles to stop MSBuild nodes dying of memory pressure [CI]	2026-10-21 15:12:58 +08:00
justme	f2658b24f7	ci: read the built binaries from an OCI archive instead of mounting the image [CI] ... buildah unshare fails on the CI host with memfd_create: Invalid argument, so extract-node-binaries.sh failed right after a successful native build. Exporting an OCI archive needs no mount (the toolchain archive already works that way); extract-oci-files.py reads the files from the layers, later layers win and whiteouts delete.	2026-10-21 15:07:46 +08:00
justme	bfcac805af	ci: do not count unreaped zombies as live processes in the guard tests [CI] ... The CI container has no init to reap killed processes, so kill(pid, 0) still succeeded for a process the guard had already killed.	2026-10-21 14:55:23 +08:00
justme	5d75fe292a	ci: take one image id when an image carries several tags [CI] ... Identical restore images accumulate one tag per commit, and buildah images <name> prints a row per tag, so the single-id check in build-linux-agent.sh failed for every commit after the first. moshiro_image_id returns one id and a test covers the multi-tag case.	2026-10-21 14:52:28 +08:00
justme	4f5a4aa3a9	ci: stop the whole pipeline on the first failure and report why the restore image is missing [CI] ... Every Linux step now runs under ci/guard-step.py. The first step that fails, is cancelled or times out marks the pipeline; running steps get SIGTERM (so traps release flock and Buildah state) then SIGKILL, steps that have not started exit at once, and any process a step leaves behind is killed. A pipeline deadline is the last resort. build-linux-agent.sh now prints the lookup error, store configuration and the restore images present when the prepared restore image cannot be found.	2026-10-21 14:48:40 +08:00
justme	13c16ad9c7	ci: keep build cache cleanup best-effort in the restore step [CI] ... The restore step failed after restoring every dependency because the final cache cleanup (buildah unshare reporting memfd_create) was not guarded. Cleanup failures now only warn, and a test requires every cleanup call to be guarded.	2026-10-21 14:41:28 +08:00
justme	c6d0c8637e	ci: mount the NuGet cache in Agent restore stages [CI]	2026-10-21 13:54:17 +08:00
justme	59c98e6d7b	ci: persist NuGet restore cache across Linux builds [CI]	2026-10-21 13:50:43 +08:00
justme	f2af7950e7	ci: persist the prepared NativeAOT toolchain archive [CI]	2026-10-21 13:46:07 +08:00
justme	a867c9f7f6	ci: pin Linux build RIDs across stages and retry runtime fetches [CI]	2026-10-21 13:40:21 +08:00
justme	806493567b	ci: share Linux restore artifacts before builds [CI]	2026-10-21 13:31:07 +08:00
justme	a6f9759238	ci: build musl only for arm64 and reuse toolchain cache [CI]	2026-10-21 12:41:58 +08:00
justme	490b008a57	scripts: unify Linux native builds and retain restore caches [CI] ... Build musl and glibc for amd64 and arm64 through one native package path. Reuse musl outputs for OCI images and OpenWrt installers, and export binaries inside a rootless unshare session. Restore the executable dependency graph into a shared persistent NuGet cache, build managed dependencies once, and run isolated native publishes in parallel. Clean temporary outputs and expired image layers under store locks, retire inactive legacy stores after seven days, and retain two toolchain exports. Add regression coverage for staging, extraction, packaging and cleanup boundaries.	2026-10-21 12:29:00 +08:00
justme	4515d7b075	scripts: publish unified Agent artifacts and restore CI coverage [CI] ... Remove Probe build jobs and release paths. Transfer Gitea publishing credentials to both Agent architecture jobs, publish each platform archive once, and run slow tests after development images are published.	2026-10-21 11:56:06 +08:00
justme	c9beeee8e3	docs: describe the unified Agent topology and link controls ... Align architecture, protocol, API, storage, security and deployment documentation with default mutual measurements and the Agent link board.	2026-10-21 11:55:55 +08:00
justme	15f0023614	web: unify Agent management and add a draggable link board ... Replace Probe pages with directed Agent links, shared deployment controls and unified display data. Add node dragging, canvas pan and zoom, saved layouts, and per-direction link controls.	2026-10-21 11:55:50 +08:00
justme	016dbc2bbf	scripts: consolidate Agent deployment and remove Probe packages ... Use Agent for native, container, systemd and OpenWrt installations and signed updates. Remove retired Probe deployment templates and smoke scripts.	2026-10-21 11:55:41 +08:00
justme	ad5efc62ff	agent: unify listening and dialing with directed mutual measurements ... Replace Probe with Agent across node runtime, Fleet contracts, backend state, API and storage. Report incoming sessions and outgoing link health through each Agent uplink, and retain directed link controls and retry. Update protocol fixtures and end-to-end coverage for the unified model; accept requested cancellation during the Fleet timeout test cleanup.	2026-10-21 11:55:36 +08:00
justme	d5ebf29e57	docs: tighten wording, soften node capability claims, link architecture to code ... - Remove the node capability paragraph from README; soften "pure collector" and "no self-update" statements in docs, AGENTS.md and READMEs - architecture.md: drop scalability goals, design decisions and open items; add a component-to-code table and code links - Replace rhetorical contrasts in security docs with direct statements - AGENTS.md: add project mental model, change map and writing style rules	2026-10-21 06:52:01 +08:00
justme	a9bf849697	docs: rewrite README and AGENTS.md etc..	2026-10-21 06:28:19 +08:00
justme	cc64df3e70	[CI] Keep line cards visible and time windows current; add 24h to 96h ranges	2026-10-20 20:44:04 +08:00
justme	a7ebf61551	[CI] [SKIP TESTS] Add detailed Probe-Agent connection diagnostics	2026-10-20 19:12:11 +08:00
justme	0fd38b2367	fix: prevent QUIC heartbeat stalls and isolate TCP retransmission counters [CI] [SKIP TESTS]	2026-10-20 18:23:31 +08:00
justme	0066a947da	fix: correct OpenWrt API metadata and earliest sample queries [CI]	2026-10-20 14:56:39 +08:00
justme	cd193378a6	probe: add native OpenWrt deployment and signed automatic updates [CI]	2026-10-20 14:15:01 +08:00
justme	24ba05fbb6	perf(ci): pin staged restore mtimes so the restore layer caches, drop build intermediates from the layer [CI]	2026-10-19 14:05:22 +08:00
justme	5c00ec434e	fix(ci): read the commit status from the public Gitea address [CI]	2026-10-19 14:03:46 +08:00
justme	1ccbb847b5	ci: build the dev images in parallel with the checks and gate only the push [CI]	2026-10-19 13:58:21 +08:00
justme	b33e50696d	fix(api): never cache anchor=latest windows as immutable [CI]	2026-10-19 11:29:34 +08:00
justme	fb9dc155f5	fix(ci): avoid leaking cache lock to MSBuild nodes [CI]	2026-10-19 11:15:19 +08:00
justme	5594d1cc86	fix(query): a connected line is always anchored at now, whatever the stores have caught up to [CI]	2026-10-19 10:48:38 +08:00
justme	e1ec7c9b04	fix(query): anchor=latest follows the newest sample of any level, not just the 1m rows [CI]	2026-10-19 10:34:04 +08:00
justme	d95acc6575	refactor(link): bound every teardown and handler wait behind one helper [CI]	2026-10-19 10:07:36 +08:00
justme	c05842fd91	fix(web): time out API requests and flag line strips that failed to refresh [CI]	2026-10-19 09:31:57 +08:00
justme	09b4630953	fix(probe): rebuild the Agent session when the datagram path goes silent [CI]	2026-10-19 09:28:29 +08:00
justme	bbbf0a70bd	chore(ci): rebuild the dev images after the CI host restart [CI]	2026-10-19 08:31:41 +08:00
justme	57ae0752aa	fix(probe): bound every step of tearing down an Agent session [CI]	2026-10-18 19:42:43 +08:00
justme	f9be98e801	fix(hosting): exit on our own when the orderly shutdown hangs [CI]	2026-10-18 10:22:34 +08:00
justme	839c45c7d4	fix(probe): end an Agent session that has gone silent and redial [CI]	2026-10-18 10:07:52 +08:00
justme	1a3feddc9d	fix(link): actually bound the server connect with a timeout [CI]	2026-10-18 03:30:43 +08:00
justme	68c7113adf	feat(update): update the backend like nodes via quadlet restart [CI]	2026-10-18 00:36:38 +08:00
justme	34190f5a04	fix(link): bound the server connect and alternate address families on retry [CI]	2026-10-18 00:23:48 +08:00
justme	0d532caf08	fix(nodes): locate agents behind a proxy and retry missing countries [CI]	2026-10-18 00:06:39 +08:00
justme	1145a53004	style(display): refine overview line layout [CI]	2026-10-12 02:21:56 +08:00
justme	52059b04c1	fix(api): refresh node build versions from heartbeat [CI]	2026-10-12 01:57:50 +08:00
justme	fe049c97fa	fix(link): reconnect after RPC send failures [CI]	2026-10-12 00:21:12 +08:00
justme	6cf6b23585	feat(ops): upload node logs and task traffic [CI]	2026-10-11 23:57:13 +08:00
justme	e82b165eb0	fix(quic): exclude locally rejected datagrams [CI]	2026-10-11 19:58:36 +08:00
justme	f6e477fb10	chore(netdiag): add standalone UDP path diagnostics	2026-10-11 19:58:35 +08:00
justme	be97e24693	fix(link): recover stale websocket sessions [CI]	2026-10-11 19:54:14 +08:00
justme	b5e275c51f	fix(quic): handle Linux dual-stack listeners [CI] ... Signed-off-by: JustMe <mr.toing@outlook.com>	2026-10-11 04:45:27 +08:00
justme	b9ec6d80b0	fix(quic): merge static MsQuic support with latest main [CI]	2026-10-11 01:05:21 +08:00
justme	a44895d860	fix(quic): statically link MsQuic for Alpine nodes [CI]	2026-10-11 01:00:37 +08:00
justme	d7b48c068c	ci: speed up backend build and parallelize image publishing [CI]	2026-10-10 07:29:11 -07:00
justme	04f116e092	ci: clean up backend test processes on cancel [CI]	2026-10-10 06:48:20 -07:00
justme	bf3481ebae	ci: remove slow tests from dev image pipeline [CI]	2026-10-10 06:25:15 -07:00
justme	b46ba69aa0	ci: raise Buildah nofile limit for Zig AOT [CI]	2026-10-10 06:09:17 -07:00