Telemetry backlog: rows since the last successful push, stored on the card, sent in full, reported when it grows

未关闭
  • 哪里不对？
#10 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
25/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
arduino, cpp
领域
data-engineering, embedded-iot

调研方向

Start with Okapi.cpp:255–296 to trace afterLogEvent() and the current push flow, then read NW-Status/QUEUE.md requisite 6 for the Feather handshake dependency. The proposal also names the status file, data file, and LED grid plan in NW-Status; inspect those entry points before estimating the changes. Done means the last-success marker survives resets, pushes cover the backlog within the Boron session limit, and the backlog is reported and recorded as described.

由索引模型根据 Issue 内容生成。

描述

enhancement

Keep a telemetry backlog: the rows logged since the last push that succeeded, kept where a reset cannot lose it, sent in full at the next push, and reported when it grows.

What the firmware does today (Okapi.cpp:255–296). _logCount counts rows in RAM and afterLogEvent() pushes when it reaches _logCountPush (5) and the board is on main power. The push prints the last five rows by SD index, sets _logCount = 0, and then waits up to 180 s for FeatherGPIO to fall. Three consequences:

  1. The counter is zeroed before the result is known, so a push that times out loses its five rows from telemetry.
  2. _logCount and _lastSdIndex are RAM. A reset (watchdog, brown-out, battery swap) forgets both, and the next push sends five rows counted from the reset, not from the last success.
  3. On backup power the push is skipped but the counter keeps climbing; when main power returns the push still sends five rows, and everything before them is never sent.

There is no backlog, only a five-row window, and the window drops rows in each of these cases.

Proposed.

  • The marker, not the count, is the state. Store the timestamp (or SD index) of the last row that a successful push covered. Write it once per success, after FeatherGPIO falls, never before. The backlog is then every row after the marker, and it needs no counter of its own.
  • Where it lives. On the card, as a line in the status file, which the logger already writes and which survives a reset at no wear cost. EEPROM would work for the marker alone (one write per successful push) but a per-row counter in EEPROM would wear a cell out in months at one row a minute, which is why the count is derived and the marker is stored. At boot, read the marker and count rows after it in the data file.
  • The push sends the backlog, from the marker to the newest row, not a fixed five, so a return from backup power or a reset sends what was missed. Cap a single push at a size the Boron session can carry and let the next push continue.
  • Report it. A TELEMETRY_BEHIND message (CAUGHT) when the backlog passes twice _logCountPush, cleared by a success; a column in the data file with the backlog at the time of the row, so the gap is visible in the record afterwards. On the LED grid it is the telemetry column (B1 yellow), per the grid plan in NW-Status.

Depends on separating "connected" from "sent" in the Feather handshake (requisite 6 in NW-Status/QUEUE.md), because a success must mean the rows arrived, not that the modem woke.

Should it survive a boot cycle? Yes. The rows it counts are on the card and survive; a backlog that forgets them while the rows remain unsent is wrong by construction.

🤖 Generated with Claude Code

https://claude.ai/code/session_01BazaY6meA6yMRDeuUsv1Wd

主要语言
C++
星标
1
派生
0
30 天内没有已合并 PR

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

NorthernWidget/Okapi_Library 的其他 Issue

查看 NorthernWidget/Okapi_Library 的全部 Issue

相似的 Issue

更多 C++ Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。