IMD SolutionComputing LibraryShop main store

Troubleshooting

Hold an Ignition Edge connection up through wireless dropouts

Edge computing and IIoT Published 6 min read

Short answer

The Gateway Network fault budget is the ping interval multiplied by the missed-ping count: a 10 s interval with 60 missed pings gives the ten-minute window you want, but only if no lower timer - the fixed 30,000 ms WebSocket idle timeout on Ignition 8.1.2, a firewall state expiry, a radio reset - closes the TCP session first. If the path keeps resetting it, move to MQTT store-and-forward.

On this page
  1. Why did raising missed pings from 30 to 60 change nothing?
  2. How do you turn 10 minutes of tolerance into ping settings?
  3. Why does the connection fault before the ping budget runs out?
  4. Do alarm delays or Edge history buffering hide the fault?
  5. Is the Edge-to-central connection MQTT underneath?
  6. When should you move to MQTT store-and-forward?
  7. How do you verify the fix with a controlled outage?
  8. Questions asked

An Ignition Gateway Network connection faults when its missed-ping counter runs out, and that counter only governs a session that is still open. The effective budget is the ping interval multiplied by the missed-ping count, so going from 30 to 60 pings bought 30 more intervals and nothing else. If the link still faults after minutes of dropout, either the interval is short or something below the gateway - the WebSocket idle timeout, a firewall state expiry, a radio re-association - closes the TCP session first. Fix the arithmetic, clear the timers underneath, verify against a ten-minute target, and if the path keeps resetting the session, move the data path to MQTT store-and-forward.

Why did raising missed pings from 30 to 60 change nothing? #

Because the tolerance you added is 30 ping intervals, not 30 minutes: the fault budget is the ping interval multiplied by the missed-ping count, and any reset below the Gateway Network closes the session before the counter can finish.

Three things produce exactly this symptom:

  • A short interval. At a 1 s ping interval, 60 missed pings is a 60 s budget. Doubling the count moved the tolerance from 30 s to 60 s, which is invisible against outages measured in minutes.
  • A session torn down from below. A re-associating radio, a stateful firewall or NAT flushing its table, or a VPN re-key closes the TCP session outright. The gateway sees a closed connection immediately, and the missed-ping count never gets to run.
  • The wrong definition edited. The outgoing connection on the Edge gateway and the incoming connection on the central gateway each hold their own ping settings, and either side can declare the fault.

The counter is the right knob only for the first case.

How do you turn 10 minutes of tolerance into ping settings? #

Divide the target in seconds by the ping interval in seconds - required missed pings = 600 ÷ interval - then write that count into both the outgoing connection on the Edge gateway and the incoming definition on the central gateway.

The connection's ping settings control how often a ping is sent and how long the response is waited for. The 8.1 manual documents the response wait, the setting named Ping Timeout, with a default of 300 ms, and every expired wait counts as one missed ping. Those waits ride on top of the budget: 60 pings × 0.3 s = 18 s worst case, small next to the target.

Worked example, with illustrative intervals:

  • Interval 1,000 ms (1 s): required count = 600 missed pings. The current setting of 60 delivers 60 × 1 s = 60 s.
  • Interval 10,000 ms (10 s): required count = 60 missed pings - the number already configured, for a budget of 60 × 10 s, the full ten minutes.

Decision: set the interval to 10,000 ms and leave the count at 60. That single interval change turns the existing count into the 10-minute budget, and 10 s stays clear of the 30 s ceiling described next.

Why does the connection fault before the ping budget runs out? #

On Ignition 8.12 the shortest timer is the fixed WebSocket idle timeout of 30,000 ms, which rejects ping intervals above 30 s; on 8.1.3 and later it is whichever you leave lowest among the Websocket Idle setting, a firewall state timeout, or the radio itself.

  • Ignition 8.1.2: a WebSocket idle timeout fixed at 30,000 ms means any configured ping interval above 30 s trips the idle timeout first. Keep the interval at 10 s and you stay below it.
  • Ignition 8.1.3 and later: the Websocket Idle timeout is configurable. Raise it above your ping interval before you lengthen the interval.
  • Firewall, NAT, or VPN on the path: read the idle or session timeout from each device. Anything below the target closes the session before your budget is spent; set 900 s to leave reconnect margin.
  • Radio pair: re-association and roaming can reset the session regardless of every software timer. A capture on the Edge host NIC that shows a TCP RST or FIN before the budget is spent proves the transport closed first, and no ping value fixes that.

Do alarm delays or Edge history buffering hide the fault? #

No - the quality overlay follows the connection fault state, so alarm delays only postpone notification and Edge history buffering only protects the archive; live tag quality still drops for the whole outage.

Approach Stops the fault Removes the quality overlay
Raise the missed-ping budget Only if the session survives the outage Only while the connection stays up
Extend alarm delays No No
Edge history buffering No No - it protects the archive and backfills later
MQTT store-and-forward No persistent session exists to fault Depends on how the central side models the data

Alarm delays are the wrong tool by construction: the overlay is produced by the connection fault, not by the alarm pipeline. History buffering is doing its job - the archive survives and backfills on reconnect - but live values read bad quality for as long as the connection is faulted.

Is the Edge-to-central connection MQTT underneath? #

No - a Gateway Network connection is its own protocol carried on a TCP session, and MQTT is only in play when the Cirrus Link MQTT modules are deliberately installed and configured.

The assumption matters because it makes store-and-forward look free. It is not: the Gateway Network connection faults as a unit, and only history survives the gap. Multi-outage buffering of the data path is a different architecture you have to build.

When should you move to MQTT store-and-forward? #

Move when dropouts are routine rather than exceptional, when data must keep arriving across outages with no persistent session left to fault, or when the timers on the transport path cannot be raised above your target.

With the Cirrus Link modules - MQTT Engine on the central gateway, a transmission module on the Edge side - published data is held at the edge until the broker accepts it, so a dropped link delays delivery instead of faulting a connection. If the data is slow-moving records rather than live values, that is the same trade-off as choosing an edge data path for periodic records. What the central side sees after a reconnect is covered in MQTT store and forward backfill after a connection outage, and fixing a missed transaction group in MQTT Engine store-and-forward covers the failure mode to watch on that path.

Decision: with the interval at 10 s, the count at 60, Websocket Idle raised on 8.1.3 and later, and firewall state at 900 s, a connection that still faults inside the target is being reset by the path - stop tuning and move the data path.

How do you verify the fix with a controlled outage? #

Block the wireless path for 480 s - under the ten-minute target - then restore it and require the connection up on both gateways, tag quality good, and no fault logged.

  1. Disable the radio, or the firewall rule that permits the Gateway Network port, for 480 s (8 minutes).
  2. Restore the path. The Gateway Network status page on both gateways must show the connection up, tags must read good quality, and the logs must show no fault.
  3. Repeat with a 660 s (11 minute) block: the fault and the quality overlay should now appear, then clear on reconnect, with Edge history backfilling the gap.
  4. If the 480 s window still faults, capture on the Edge host NIC during the block. A TCP RST or FIN arriving before the budget is spent names the transport as the closer, and the device timers two sections up are the next thing to change.

Questions asked

The connection faults at almost the same duration every outage - what does that tell you?

A constant duration means a fixed timer is firing, not a random radio event. Measure it: roughly 30 s points at the WebSocket idle timeout on Ignition 8.1.2, while a longer constant points at a firewall, NAT, or VPN state timeout somewhere on the path. A genuine missed-ping fault instead lands near your configured interval multiplied by the count.

Can the missed-ping count go above 60?

Yes. Nothing in the ping mechanism stops at 60 - a ten-minute tolerance at a 1 s interval needs a count of 600. The practical ceiling comes from the WebSocket idle timeout and any stateful device between the gateways, not from the field itself. A 10 s interval keeps the count at 60 for the same budget.

What do you give up by raising the ping interval?

Detection speed. The interval times count product is both your dropout tolerance and your dead-gateway detection time: at 10 s times 60, a gateway that is genuinely unreachable shows the quality overlay only after the full budget. That is the trade you are accepting, so pick the product deliberately rather than the interval alone.

Do both gateways need the same ping settings?

Yes, and each side holds its own definition: the Edge gateway's outgoing connection and the central gateway's incoming connection have separate ping settings, and either counter can declare the fault. Apply the same ten-minute budget to both, save, then restart the connection so each end re-reads the values.

What does a TCP RST in a packet capture during an outage prove?

That something between the gateways actively closed the session - a re-associating radio, a firewall flushing state, or the host stack - so the missed-ping counter never had a chance to matter. No ping setting prevents that fault. Lengthen the state timeouts on the offending device, or move the data path to MQTT store-and-forward.

Keep reading

  • Comparison

    Getting Wonderware Historian data into Ignition Edge for MQTT

    Yes — Ignition Edge can publish this data as MQTT Sparkplug, but through two separate routes. Live values come from the existing DAServer acquisition layer behind a DA-to-UA gateway, because Edge's OPC client is UA-only. History and aggregates come from SQL queries against the Historian Runtime database, using its documented stored procedures.

    Edge computing and IIoT

  • How-to

    Bridging OPC UA PLC data to MQTT with an edge PC

    Keep the OPC UA server on the PLC side and give MQTT publishing its own role on a separate edge PC that acts as the OPC UA client and MQTT publisher. Commission the two links independently — first a verified tag read, then a verified subscriber message — and only then widen the tag set.

    Edge computing and IIoT

  • Selection guide

    Choosing an edge data path for periodic records

    For independent periodic samples, Tag Historian with Edge Sync Services is the reliable path and buffers 35 days locally. For one atomic row per collection cycle, use MQTT Transmission records landing in MQTT Recorder. For true transaction groups, license a standard gateway for that node.

    Edge computing and IIoT