How should I reconcile one completion with two saved rows?

LiamBrooks0838 · 2025年1月28日 11:02 UTC

已关闭

主帖中文翻译暂不可用,以下显示英文原文;回复保留原文。 查看英文原文

LI
LiamBrooks0838
I maintain an FR10 teaching bench with a PLC job interface. We unplugged the monitoring laptop after a sample check had completed. Reconnecting added another completed row to the lesson record, although nobody submitted another job. The PLC completion state stayed high throughout. Our monitor treats the first high value it reads as an event. I need a recovery rule that an instructor can explain and support without clearing evidence or submitting the check again.

15 条回复

LU
LucaBarnes0526

Use the job reference to recognise the same result, provided that reference survives the reconnect too.

22分
LI
LiamBrooks0838

The PLC retains a job counter. The monitor stores it in each row, but its duplicate check lives only in memory. Both saved rows have the same counter.

2分
LU
LucaBarnes0526

Then you've got something firmer than the time column to compare; does that counter ever start again?

25分
RO
RosaChan1075

We got caught by that with a test logger. A counter was unique until the equipment restarted. What did your PLC do during this particular outage: remain running, or restart as well?

4分
LU
LucaBarnes0526

Good distinction, I was assuming only the laptop had gone away.

23分
LI
LiamBrooks0838

Only the laptop disconnected in this test. The PLC kept running. Its counter can be reset during maintenance, though, so it is not a permanent unique identifier by itself.

17分
LU
LucaBarnes0526

I'd mark this pair as one known duplicated result and leave the wider restart case explicitly unfinished.

9分
VI
VictorChen1142

Please include the operator view. Two rows corrected in the database can still leave an instructor with a misleading total on screen. One explanation for both (and somebody responsible for making the correction).

19分
TO
TobyCarter1014

What prevents a later legitimate job with a reused counter from being silently discarded by the proposed fix?

-1分
LU
LucaBarnes0526

Nothing in my first suggestion covered that, you're right; the counter needs a run identity or an equivalent documented boundary.

16分
LI
LiamBrooks0838

I have added a proposed controller-run identity to the interface review. For the existing data, only the observed reconnect pair is being reconciled. Unknown matches will be flagged for review instead of discarded.

16分
RO
RosaChan1075

Also try losing the connection after the row is saved but before the monitor records that it handled it. That's where our logger duplicated things even with a good job identifier.

7分
LI
LiamBrooks0838

That reproduces the problem too. We are changing persistence so recognising an already recorded result does not depend on the monitor reaching a later in-memory step. No new work is submitted during recovery.

7分
LI
LiamBrooks0838

The laptop reconnect cases now display one completion in our offline tests. Maintenance reset handling still needs the PLC owner's review, so I am keeping the issue open on the handover sheet.

16分
LU
LucaBarnes0526

Does the instructor see the unresolved state after a counter reset, or is that still only in your log?

11分

讨论已关闭

此讨论已连续六个月没有活动,现已关闭新回复。 最后活动: .