IMDHMIControls LibraryShop main store

Troubleshooting

Why PLC symbol changes break SCADA tags in OFS

SCADA and data logging Published 5 min read

Short answer

The tags drop because OFS resolves symbolic item names through a symbol source that no longer matches the downloaded PLC application. Replace the stale XVM symbol file with a fresh export, or move to OFS 3.63 SP3 with "Using Data Dictionary", "No Communications Break" and the PLC build's preload option so the server reads the updated dictionary before scanning.

On this page
  1. How the symbol source goes stale
  2. Checks that separate stale symbols from a communication fault
  3. Restoring access with a fresh XVM export
  4. Making the change survive the next download
  5. Tracing the failure on a five-PLC line after three downloads
  6. Questions asked

Your tags drop because OFS resolves symbolic item names against symbol information that no longer matches the application running in the PLC. The fix that restored access in the reported installation was regenerating the XVM symbol file and replacing the old one in the server configuration. The fix that stops it recurring is OFS 3.63 SP3 or later reading the embedded Data Dictionary, with "Using Data Dictionary" and "No Communications Break" ticked in the server and "Preload on build changes" ticked in the PLC project's embedded data settings. The case below is that installation: OPC Factory Server and Vijeo Citect 2015 on one computer, five PLCs on Ethernet, repeated downloads to one of them.

How the symbol source goes stale #

OFS translates a symbolic item name into PLC data access using symbol information loaded when its configuration started. That information comes from one of two sources: an XVM file exported from Unity Pro and associated with the device definition, or the Data Dictionary embedded in the controller application. A download that changes the application can change the symbol information on the PLC side while an exported XVM file sits unchanged on the server, and from that moment the server's mapping no longer matches what the PLC will answer. The SCADA keeps its tag entries and loses the values behind them.

The interruption itself is documented behavior, not a fault to chase: Schneider's FAQ on the symptom states that a build that modifies the data dictionary forces the server to reload it, and OPC clients temporarily lose synchronization while that happens. Symbolic access remains the right approach, because it is what lets a server read controller data without mapped memory addresses, the approach described in reading M580 I/O in Plant SCADA without mapped memory addresses.

Checks that separate stale symbols from a communication fault #

Work through these readings before changing anything:

  • Tags from one PLC fail while the other four read normally: the shared server process, the shared computer and the network are all serving good data, so the fault sits between that PLC's downloaded application and its symbol source. Go straight to the XVM or dictionary fix.
  • Every item of one PLC fails, including tags you never changed: read the OFS diagnostics for that device first, because device communication and symbol loading both need examining before you replace any symbol file.
  • Tags from several PLCs fail together: stop, the fault is shared rather than symbolic. Check the OFS service, the SCADA machine and the network first.
  • The SCADA tag entry itself is gone from the project: that is a configuration or deployment change on the SCADA side, not an OFS item-resolution problem.
  • A refreshed XVM restores access: confirmation that the previous file no longer matched the application's symbols.

Because OFS and the SCADA share one computer in this installation, a read that succeeds inside OFS is the sharpest test available: it clears the PLC-side lookup and leaves the SCADA address or client connection as the only remaining suspects.

Restoring access with a fresh XVM export #

This is the fix that worked in the reported installation, and it needs no upgrade to the installed software:

  1. In Unity Pro, select the variables, right-click, and export them in XVM format.
  2. In the OFS Configuration Tool, associate the .XVM file with the device definition of the affected PLC.
  3. Confirm the path the server actually loads. A new export changes nothing if OFS keeps reading another file from another location.
  4. Restart OFS and confirm the failing items return with good values, together with the four untouched PLCs.

Schedule this for a window where a short SCADA outage is acceptable: restarting the server drops every client's items, and if these tags carry alarms, the alarm display goes quiet with them.

Making the change survive the next download #

The XVM export is manual and goes stale again at the next download that touches symbols. The embedded Data Dictionary route removes the manual step, and it needs three settings working together, of which the reported installation had two:

  • OFS at version 3.63 SP3 or later, the release that reads the dictionary stored in the PLC.
  • In OFS, "Using Data Dictionary" and "No Communications Break" ticked. Both were already ticked in the reported case, and on their own they did not stop the tag loss.
  • In the PLC project settings, under embedded data settings, "Preload on build changes" ticked, with the Unity Pro build generating the dictionary into PLC memory. This was the missing piece: it lets the server preload the updated dictionary before starting the new scan, instead of scanning a changed application against old symbol information.

With all three in place the server fetches the updated dictionary from the controller, scans against it, and no manual export is needed. Reading the dictionary from the controller also removes the failure mode behind resolving M580 OFS variable address drift with the Data Dictionary, where added variables shift the addresses behind unchanged names.

Tracing the failure on a five-PLC line after three downloads #

Inputs, all from the reported installation: five PLCs (PLC1 to PLC5) on Ethernet; one computer running OFS and Vijeo Citect 2015; three downloads of changes to PLC1 before the tag loss was observed; zero downloads to the other four; "Using Data Dictionary" and "No Communications Break" ticked; "Preload on build changes" unticked.

The rule in words: count the PLCs whose SCADA tags fail and compare that set with the set whose application was downloaded. The stale-symbol diagnosis holds only when the two sets are the same single PLC.

The arithmetic:

  • PLCs configured: 5
  • PLCs downloaded: 1 (PLC1)
  • PLCs still reading after the third download: 5 − 1 = 4
  • PLCs observed failing: 1 (PLC1)

The failing set and the changed set are the same PLC, so the four good PLCs clear the shared suspects (server process, computer, Ethernet) and leave PLC1's symbol source as the cause. Treat the number three as this site's reproduction count, not a threshold: the mapping can break on the first download that changes symbols, so check the affected tags after every download rather than waiting for the third.

The decisions: one changed PLC failing means you fix the symbol source, regenerate the XVM now and enable the dictionary preload before the next download. Two or more PLCs failing together means you stop and investigate the shared OFS and SCADA machine first, because no stale symbol file produces that pattern.

Questions asked

Why did the tags disappear only after the third upload?

The count is a reproduction detail of that installation, not an OFS threshold. Every download that changes the symbol database can invalidate the mapping on the first one; nothing waits for a third. What the three-upload sequence proves is only that the loss follows changes to that PLC. Inspect the affected tags after every download and you will catch it on the first.

Do the other four PLCs lose data when I replace the XVM file?

Yes, briefly. OFS is one server process, so loading a new configuration reconnects every device it serves, and all five PLCs re-establish their items when it comes back. Schedule the restart for a window where a short SCADA outage is acceptable, and verify after the restart that counts and values from the four untouched PLCs recover along with the changed one.

Is the XVM export still needed once the Data Dictionary route works?

No. When the PLC build stores the embedded dictionary in controller memory and OFS 3.63 SP3 or later reads it with the preload options set, symbolic items resolve from the controller and manual exporting stops entirely. Keep the export procedure documented anyway: it is the fallback for any OFS installation that has not been upgraded, and it restores access the same day a version rollback happens.

What do the OPC clients see while the dictionary reloads?

A temporary loss of synchronization. When a build modifies the data dictionary, OFS has to reload it, and clients such as the SCADA lose sync with the affected items until the reload completes. Without preloading that window sits after the PLC accepts the new application; with "No Communications Break" and preload on build changes enabled, OFS fetches the updated dictionary before it starts the new scan, so clients keep their values.

Keep reading

  • Troubleshooting

    Reading M580 I/O in Plant SCADA without mapped memory addresses

    Bad I/O on %Q1.1.20 means Plant SCADA is reading a topological M580 channel address that no driver can resolve as a register, and no Modbus table exists to look up. Serve the points through the OFS gateway using PLC tag names, with an XVM export now and the Data Dictionary permanently, and add the missing variables by online change so the 24/7 line never stops.

    SCADA and data logging

  • How-to

    Serve generic Modbus devices to SCADA through OFS with the PL7 driver

    OFS does serve generic Modbus devices: select the PL7 driver in the device alias, not Unity, because Unity issues Schneider-extended function codes a standard slave rejects while PL7 issues the public read/write set. Import the tag definitions as a .SCY file, make OFS the only Modbus master on the link, and point Citect and every OPC DA or UA client at that one server.

    SCADA and data logging

  • Troubleshooting

    Fixing CODESYS OPC DA memory growth stopping SCADA reads

    The reads stop because WinCoDeSysOPC.exe, a 32-bit OPC DA server process, climbs to its 2 GB memory ceiling over weeks and then stops serving tags; killing it only resets the clock. The resolution that held was installing the current, licensed CODESYS OPC DA server — 3.5.22.40 in place of the CODESYSV35SP10Patch2 install — and proving it through the old failure window.

    SCADA and data logging