Repository navigation
Release v1.3.70
Release v1.3.70
v1.3.70 follows another review of the server, the web app and the design. It fixes a v1.3.69 regression that froze the admin page after a server restart, writes about a twelfth as much to disk, and fixes a set of display problems.
Fixed (a v1.3.69 regression)
- After a server restart (each upgrade), or when a session ended while the page was open, the admin page and a private dashboard opened with the admin session stayed frozen on their last values. The refused stream ticket went unnoticed. They go to the sign-in page again, as before v1.3.69.
Storage
- The agents' live numbers are kept in memory and written once a minute, and also when a server goes offline, before a backup and at shutdown.
- TCPing samples are still written before a push is answered, now in groups of up to 2 s.
- Resource minutes are written every 10 minutes.
- Measured on a copy of 131 servers, side by side with v1.3.69: 236.6 → 18.9 KB/s, or 20.9 → 1.7 GB a day, with the same CPU.
- A clean stop loses nothing. A crash or a power cut can lose the last minute's "last seen" times and up to 10 minutes of resource minutes, but never TCPing history.
- Storage failures are still caught: while writes fail, pushes are refused and /healthz answers 503.
PULSE_LIVE_FLUSH=0restores the previous writes.
Security
- Setting the password after a damaged database now needs the one-time setup code. The server prints it in its log at start and writes it to
setup-code.txtbeside the database; the sign-in page asks for it. v1.3.69 accepted the server's own address instead, but behind a reverse proxy on the same machine every visitor has that address. - Three more agent endpoints answer an unknown server ID exactly as they answer a wrong secret: TCPing results, the legacy ingest, and registration without a secret.
- The TCPing interval cannot be set below 5 s. At 1 s, 137 servers would add about 13 GB of history a day.
Dashboard
- 「运行时间」 no longer breaks onto two lines in the 1024–1119 px table.
- Network speeds of 100 MB/s and more no longer run into the uptime column: they show one decimal from 100, and the next unit before 1000.
- Phones: a lost connection is shown on the line under the title. The list no longer jumps down and back.
- English: switching to Latency no longer wraps the toolbar at 1024–1079 px.
- A long site name stays on one line in the navbar.
- Glass: the 资源 | 延迟 switches and the page buttons are frosted like the filter beside them.
- Cards: a server without a flag or OS logo gets that room back for its name.
- Speeds and uptimes use even-width figures at every width.
- Look:
- sorted and hovered column headers use the menus' accent;
- the footer lines up with the cards' content;
- the chart tooltip has a shadow again.
Server details
- A TCPing probe answered between lost ones is drawn as a dot. It was invisible, yet it set the axis and the average.
- An open chart now shows a stretch where every probe fails as it happens; it used to stand still until the first answer.
- A new server's first minute and first probe are drawn, on an axis of at least 10 minutes in whole minutes.
- 资源:
- a tab with no data or a failed read shows its note alone, without four blank cards;
- averaged points name the minutes they cover (03:20–03:40);
- each tooltip line names its value in full, such as 「CPU 使用率」 or 「CPU 平均使用率」 for an average, so a short name no longer sits far from its number;
- the newest dot is no longer cut in half.
- The 资源 charts come back after an admin changes the TCPing targets, and they and the latency view refresh when a tab is shown again.
- The time range menu highlights the chosen span when opened with the pointer.
Admin and sign-in
- Errors appear in the dialogs or as notices, never as the browser's alert box. Notices sit at the bottom right from 640 px, clear of the navbar.
- The share link row matches the dialog's other controls.
- A row's second line no longer starts a wrapped line with "·".
- The settings menu stays on screen on phones.
- Error text is darker: 5.7:1 contrast instead of 3.3:1.
Also
- The resource history answers are cached: 8.1 → 0.5 ms of CPU a request.
- The admin page downloads backups compressed: 186 MB → about 22 MB for 131 servers. Scripts get them as before.
- Custom
<script>tags keep their attributes, so trackers that readdata-*(Umami, Plausible) work. - A stream cut right after its first update backs off instead of reconnecting every second.
- Switching 延迟 → 资源 during a read no longer leaves a summary refresh running.
Compatibility
- Agents are unchanged (1.3.68). Pulse-Hotaru gets the server changes; its pages keep working as they are.
- A downgrade reads the database. Settings added since are dropped when saved, and the setup lock is not enforced.
Verified
- Go, with the race detector, including:
- the live state: admin edits between a push and a flush, kill -9 and clean stops, offline marking, clock steps, storage failures;
- the setup code;
- the uniform agent answers;
- the history cache;
- the compressed backup;
- the interval bounds.
- Side by side on copies of 131 servers:
- the write rates;
- an admin rename kept through pushes, a flush and a kill -9;
- TCPing history intact after kill -9;
- a clean stop within 5 s of the last push.
- Chromium, Firefox and WebKit:
- each display fix;
- header and column fit, 768–1920 px in zh and en;
- the stream and session scenarios (restart, ended session, privacy, share link, cut, stall, offline);
- the regression suites.
📦 Docker Images
xhh1128/pulse:1.3.70xhh1128/pulse:latest
📥 Downloads
Client binaries
| Platform | File |
|---|---|
| Linux amd64 | probe-client |
| Linux arm64 | probe-client-arm64 |
| macOS Intel (amd64) | probe-client-darwin-amd64 |
| macOS Apple Silicon (arm64) | probe-client-darwin-arm64 |
| Windows amd64 | probe-client.exe |
| Install Script (Linux + macOS) | install.sh |
Server standalone binaries (uploaded by the standalone workflow)
| Platform | File |
|---|---|
| Linux amd64 | pulse-server-standalone-linux-amd64 |
| Linux arm64 | pulse-server-standalone-linux-arm64 |
Quick Install (Linux & macOS):
curl -sSL https://raw.githubusercontent.com/xhhcn/Pulse/main/client/install.sh | sudo bash -s -- --id YOUR_ID --server http://YOUR_SERVER:8008One-click migration to a new host:
PASSWORD="$OLD_ADMIN_PW" \
sudo -E ./scripts/migrate.sh --from https://old.example.com