The acknowledgement is logged before the request

TheoBell0671 · 30 Jan 2025, 16:15 UTC

Closed
TH
TheoBell0671
I'm piecing together a failed reference-plate check on our FR10 bench. In the merged export, the PLC acknowledgement appears before Python's request. Python restarted halfway through the session and reused the same job label. The clocks disagree too. Or at least their displayed times do. How much of this attempt can I honestly reconstruct before I tell the next shift what happened?

10 replies

BE
BethBennett0759
Replying to TheoBell0671

I've had merged logs create a very convincing event order that never happened. Keep each source in its own order first. Find shared transitions and preserve the original files. The repeated job label means time sorting alone won't identify the attempt.

10 points
TH
TheoBell0671
Replying to BethBennett0759

Separated them. Python has a startup line between the two uses of the label. PLC has one acknowledgement transition, but I don't yet know which request led to it.

16 points
BE
BethBennett0759
Replying to TheoBell0671

Useful boundary. Any request payload or sequence value captured on both sides?

8 points
TH
ThomasCarter0973
Replying to TheoBell0671

Could the acknowledgement already have been high when the application restarted?

11 points
BE
BethBennett0759
Replying to ThomasCarter0973

Yes, check that in the PLC trace. A high value read after restart isn't necessarily a new transition.

17 points
TH
TheoBell0671
Replying to BethBennett0759

The trace shows it went high before the application startup and stayed high. The restarted application logged its first observation as 'ack received'. That explains the duplicate acknowledgement line, though not whether the first attempt finished.

1 points
LU
LuisBrown0956
Replying to TheoBell0671

Treat that first attempt as uncertain unless a separate completion record identifies it. For the revised logging, distinguish observing existing state from receiving a new transition. Include an application-run identifier so a reused label doesn't look like one continuous attempt.

18 points
VI
VictorBrooks0794
Replying to LuisBrown0956

Put uncertain on the handover. Don't let the next shift read timeout as permission to repeat.

9 points
BE
BethBennett0759
Replying to TheoBell0671

And keep the raw timestamps alongside any estimated alignment. Someone checking your reconstruction needs to see where the evidence stops and your interpretation starts.

8 points
TH
TheoBell0671
Replying to VictorBrooks0794

Handover now says the first attempt's completion is unconfirmed and names the controls lead for reconciliation. We've retained separate source timelines. The logging change is agreed, but I can't recover a completion record that wasn't captured.

1 points

Discussion closed

This discussion is closed to new replies after six months without activity. Last activity: .