Sawmill, edger infeed. The historian has been putting four second holes in the infeed tags since Monday, about a dozen a shift, and never twice at the same time of day. Everything else on that collector is clean so I'm pretty sure it isn't the server end.
The infeed I/O hangs off a Modbus TCP gateway polling three remote blocks at 200ms. During a hole the gateway's diagnostic page shows timeouts climbing on block 2 only and the other two keep counting up fine. I tried dropping the poll to 500ms in case I was hammering the thing, and the holes came out the same length and the same number, so that idea is dead.
Before I start blaming the gateway firmware, is there anything that makes one block go quiet for four seconds and then behave itself for an hour?
Block 2 only, so look at block 2. Does its lead cross the saw anywhere?
Runs in its own duct the whole way, nowhere near the blade. Swapped it for a new lead this morning anyway and moved it to port 6 while I had the door open. Three more holes by lunch, same four seconds each.
Link light never blinks out either. I watched that port through two of them and it sat solid green the whole time, so I'm not sure a lead was ever going to be it.
Sorry if this is obvious, but does the gateway write down which block it lost, or just that a poll timed out?
Get the supplier to send another block. I reckon it's gone half alive on you, answers most of the time, and that beats a week of Wireshark.
New block was in before first break and I've had two more holes since, so that's a day gone and a spare I didn't need. No luck at all today.
To answer fee_m, yes, it logs the slave it lost, and it's the same one every time. While the door was open I noticed something though. The old block is still on the rail with 24VDC on it and its lead still in the switch. The millwrights changed it on Friday, left the old one sitting there, and never wrote it up.
There it is. It isn't dead, it is sitting there holding the same address as the new one.
Nothing polls the old block so it says almost nothing, but it still announces itself every few minutes the way any idle device does, and each time it opens its mouth the switch moves that address onto its port. Your gateway's next polls go to a box with nothing to say, you get your four seconds, then the live block talks and the table flips back. It also explains the link light, because both ports are up throughout.
Pull the lead out of the old one, better still take it off the rail. One address, one device. The flat network side of it is written up here: https://plctr.com/advanced-plc-networking-and-communication-protocols/
Got ten minutes at the blade change and took the old block off the rail, lead and all. Eleven hours of trend since and not one hole, where I'd normally have a dozen.
So it was two blocks sitting on 192.168.30.24, one live and one quietly stealing the address back every few minutes. Neither the gateway nor the Stratix called it anything, which is how I managed to spend a day on a patch lead and a block I didn't need.
I still can't tell you why it was always four seconds and never two or ten. Thanks petek, and fee_m, the question about what the gateway actually logs is what sent me back to read the log properly.
My wrong guess cost you a block. Decent spare now though.