IMD SolutionComputing LibraryShop main store

Troubleshooting

Troubleshooting a PLC agent that will not go online remotely

Remote access and virtualization Published 7 min read

Short answer

A green remote session proves only that you reached the gateway. Prove the programming connection locally first, then the agent's target address and subnet, then the agent type and protocol, then the client — one boundary at a time. Most of these faults are a mistyped target address or an agent type that carries the wrong protocol.

On this page
  1. Isolate the failing boundary first
  2. Prove the PLC connection locally
  3. Verify the target address and the subnet arithmetic
  4. Match the agent type to the programming protocol
  5. Treat Modbus as a path test, not the fix
  6. Prove the full path, then change one thing
  7. Questions asked

A PLC that will not come online through a remote access gateway is rarely a tunnel fault, and the fastest fix is almost always one of two edits: correct the agent's target address digit for digit against the PLC's configured address — for example 192.168.10.31 mistyped as 192.168.1.31, which masks to 192.168.1.0 instead of 192.168.10.0 and sends the agent's traffic to the default route — or switch to the agent type that carries the PLC vendor's programming protocol rather than data traffic on port 502. The green session in your remote client ends at the gateway — on a SiteManager, at its Uplink1 port — and proves nothing behind it. Work four checks in order: prove the programming connection locally, verify the agent's target address and subnet, test the path through the agent, then test the remote client. Most of these faults turn out to be a mistyped target address or an agent type carrying the wrong protocol, and both are fixed in minutes once you know which boundary is down.

Isolate the failing boundary first #

Run these tests in order, one change at a time:

  1. Local programming test. From a laptop on the same switch as the PLC, go online with the programming software.
  2. Target address test. Compare the agent's target address with the PLC's configured address, then run the subnet check below.
  3. Path test through the agent. Read one known value through the agent — its own status, or a single Modbus register — to prove the gateway reaches the PLC at all.
  4. Client test. From the remote engineering PC, repeat the exact connection that passed locally, same software and same settings.

Each test ends at one boundary. When a test fails, everything before it is proven good, so stop at the failing boundary and fix it before touching the next: two simultaneous changes leave you unable to say which one mattered.

What you see Boundary that is down Decision
The remote client cannot open a session to the gateway Uplink path Restore the session before any PLC work
Session open, agent shows offline Target address or gateway-to-PLC path Redo the address and subnet check; leave the agent type alone until it passes
Agent online, software will not connect Agent type or client networking Switch to the documented agent type, then test the client on a physical PC
A Modbus read works, programming still fails Protocol pairing, not the path Stop editing Modbus settings; change the agent type
Works locally, fails only through the client Client-side networking Check virtual adapters, VPN and local firewall between software and session

If the session itself will not open, the uplink is the fault — and when corporate firewalls and proxies sit in that path, remote SCADA access through firewalls and proxies is the subject to work. Everything below assumes the session opens.

Prove the PLC connection locally #

Before changing any agent setting, read the PLC's configured address and subnet mask from its project and write them down. Connect a laptop to the same switch as the PLC and go online with the programming software exactly as you would during commissioning.

If that fails, the fault is cable, switch or addressing on the local network, and a remote agent cannot repair it — every agent change you make meanwhile discards your record of what worked. Fix the local path first. If it passes, the PLC, its address and its protocol are cleared, everything from the gateway outward is now suspect, and you hold a known-good reference for every later comparison.

Verify the target address and the subnet arithmetic #

The agent is the device-side endpoint the remote client connects through. In the gateway's management pages — for a Secomea SiteManager, the agent list under GateManager → Agents — set its target to the PLC's configured address, never to the gateway's own LAN address. Then check the subnet, not just the digits.

Worked example, using illustrative addresses. The PLC project shows 192.168.10.31 with mask 255.255.255.0; the gateway's DEV1 port holds 192.168.10.5.

  • PLC network: 192.168.10.31 AND 255.255.255.0 = 192.168.10.0
  • DEV1 network: 192.168.10.5 AND 255.255.255.0 = 192.168.10.0

Both ports mask to 192.168.10.0, so the gateway delivers to the PLC directly on the LAN. Now the common failure: one transposed digit in the agent target, 192.168.1.31. That masks to 192.168.1.0, which is not 192.168.10.0, so the gateway forwards the agent's traffic to its default route instead of the PLC. The agent reports offline and no driver change will alter that result.

The decision: the agent target must equal the PLC address digit for digit, and the gateway port and the PLC must mask to the same network when they share a subnet. If the PLC sits on a different subnet, a documented route between the two is a separate check — verify it before blaming the agent.

Match the agent type to the programming protocol #

Agents differ in what they carry, and the gateway documentation draws the line plainly. Some agents give the remote client — LinkManager, in this scheme — Layer 2 access to the gateway's LAN, so the programming software addresses the PLC's own IP as though the PC sat on the local switch. The generic device agent maps only the TCP ports you configure by hand, so every port number must come from the PLC's communication documentation, never from a guess. Pre-configured agents cover common services, a WWW agent for a device web page among them; use one when it matches your service and ignore it when it does not.

A listed driver is an option, not a qualification. A Koyo driver offered by the gateway does not by itself document a programming path for a Click PLC: confirm the pairing from the gateway's agent documentation and the PLC's communication documentation together. Where neither names the pairing, commit to the agent type that bridges the client onto the gateway's LAN and let the software talk to the PLC's own address — that removes the port-mapping question entirely.

Treat Modbus as a path test, not the fix #

The advice to try Modbus is half right. Modbus TCP data traffic runs on registered port 502 and serves clients reading and writing registers, while PLC programming software on Ethernet normally speaks the vendor's programming protocol. A successful Modbus read through the agent still earns its keep: it proves the target address and the gateway-to-PLC path, which is boundary two of your four tests.

It proves nothing about the programming protocol. If the Modbus read succeeds and the programming software still will not connect, stop changing Modbus settings: the boundary that is down is the protocol pairing at the agent, not the network. Reachability mistaken for protocol support is the same trap as a link where ping works but the PLC program download fails — the session being up says nothing about the path the download needs.

Prove the full path, then change one thing #

With the session open, connect from the remote PC using exactly the method that passed locally, and confirm the PLC identity the software reports before writing anything to the controller. Never bring a PLC online remotely unless someone on site has confirmed the machine is in a safe state: the first accepted connection can energize outputs.

Keep the access inside your policy. IEC 62443 expects remote access to industrial assets to be authenticated per user, restricted to named devices, and logged — so repair the agent for this one PLC rather than widening the path until everything is reachable.

If the client test is the only failure, work the client. A virtual machine running the programming software adds a virtual adapter, and NAT mode puts address translation between the software and the session; the steps for fixing a VM remote PLC connection with a NAT adapter apply unchanged. A corporate VPN, a local firewall, or a second network adapter routing the PLC's subnet out the wrong interface does the same job of breaking a connection that works everywhere else.

Decisions by case: local test fails — fix the cable, switch or addressing on site. Agent offline — the target address or subnet, from the arithmetic above. Agent online, software offline — the agent type or the client; test on a physical PC before blaming the gateway. Modbus reads, programming fails — the protocol pairing; change the agent, not the network. Works locally, fails remotely — the client side, starting with virtual adapters and firewalls.

Questions asked

The gateway shows online in my remote client but the agent stays offline. What does that mean?

It means the gateway's own attempt to reach the address you entered is failing, which is a fault before any protocol is involved. Work it in order: compare the target with the PLC's configured address, confirm both mask to the same subnet, and confirm a route exists if they differ. Only when all three pass does the agent type become the suspect.

Can I just switch the agent to Modbus to get online?

Only for data access. Modbus TCP traffic uses registered port 502 and serves clients reading and writing registers; the programming software needs the vendor's programming protocol. Run one Modbus read as a diagnostic, since it proves the address and the path, and let it carry data meanwhile, but do not expect it to bring the programming connection online.

Where do the port numbers for a generic device agent come from?

From the two documents that own them: the PLC's communication documentation states which Ethernet port the programming protocol uses, and the gateway's agent help states what the generic agent expects. If the two do not line up, that mismatch is the answer. The pairing is not documented, so bridge the client onto the gateway LAN instead of mapping ports.

The connection works locally but fails through the remote client. What changed?

The client side. Compare it against the known-good local setup: a VPN, a local firewall, or a second network adapter sending the PLC's subnet out the wrong interface are the usual differences. Run the software on the physical PC first; if it connects there, the gateway and agent are cleared and the fault is in the client's networking.

Keep reading

  • Troubleshooting

    Ping works but PLC program download fails over remote access

    ICMP success proves only the routed IP path, not the programming session. A download opens separate query and download connections on documented TCP and UDP ports, and broadcast discovery does not cross a routed VPN. Point the controller's default gateway at the remote-access router LAN address, permit both programming ports to the controller LAN IP, and add the controller manually by IP.

    Remote access and virtualization

  • Troubleshooting

    Remote PLC access stops after a service plan change

    When remote PLC access dies after a service plan change, the cloud entitlement layer failed, not the gateway, the LAN or the PLC. Test the four layers in order — gateway power and LAN, gateway configuration, service entitlement, user access — and repair only the layer that fails; no reset, cable change or PLC edit restores a removed subscription.

    Remote access and virtualization

  • Comparison

    Remote SCADA access through firewalls and proxies

    Put the SCADA application and its data connections on the plant side of the firewall and carry only an authenticated VPN tunnel and a rendered remote session across it. A proxy forwards traffic under its administrator's policy and issues no licenses; session CALs are deployed and tracked on the session host. Restrict tunnel routes to that one host, license by counted people or devices, and keep direct port forwarding to an HMI host out of the design.

    Remote access and virtualization