IMD SolutionComputing LibraryShop main store

Troubleshooting

Troubleshooting historian trend client connection failures

Edge computing and IIoT Published 8 min read

Short answer

A trend client connection failure has two first checks. If trend pens or Historian Client still load data, compare the failing client's connection definition against a working one instead of restarting the server. If every client fails, verify the historian and SQL services, then test TCP reachability from the failing node to the configured endpoint, port 1433 only if that is the configured port.

On this page
  1. What the message proves, and what it does not
  2. Pick the first check from the scope of the failure
  3. Compare the failing client against a working one
  4. Test the server and the network path
  5. Testing TCP 1433 from a failing trend client node
  6. Correct the faulted layer without disturbing working trends
  7. When the basics pass: versions and reported defects
  8. Questions asked

The error "Connection failure. Unable to reach INSQL/Historian server" names the failed step, not the cause: a stopped service, a blocked port, a rejected login and a wrong connection definition all end at the same text. Decide the first check from the scope of the failure. If trend pens or Historian Client still retrieve historical data on the same network, compare the failing trend client's connection definition against a working one before touching the server. If every client fails, verify the historian and SQL services, then test TCP reachability from the failing node to the configured endpoint, port 1433 only when that is the port the server actually uses. This article walks both branches, works the port test through with real numbers, and ends with the version and escalation checks.

What the message proves, and what it does not #

The message reports one event: the client's attempt to establish its historical connection failed. The historian connection driver's own documentation lists the causes for that offline connection state as the server being offline, a network issue, or an incorrect username or password, the same three causes the branches below test in order. A stopped server, an unreachable SQL endpoint and a rejected login all produce identical operator-facing text, so the message identifies the failed step and nothing more.

A trend connection crosses four layers: the display's data-source selection, the client's connection definition, the network path, and the server's authentication and query processing. A failure in an earlier layer stops the client before any tag query, so a transport or login fault is not repaired by editing a pen's tag name. The reverse holds too: if the client connects and other tags load, the fault sits in the failing pen's historical tag reference, time range or display configuration, not in the connection. And a successful test on the server itself proves nothing about a remote workstation, because the two travel different network paths.

Pick the first check from the scope of the failure #

  • Trend pens or Historian Client still load historical data and only the trend client fails: compare the two clients' connection definitions side by side. Keep the working client open as a reference and leave the server alone.
  • Every client, on every node, fails: check the historian components and the SQL service behind the connection, then test TCP reachability to the configured endpoint.
  • A remote node fails while clients on the server itself work: a network-path fault. Run the port test from the failing workstation, not from the server.
  • Basic checks pass and the trouble began after a network or infrastructure change: capture exact client and server versions and check the vendor's support portal for a reported defect before changing configuration.

The first case is the strongest discriminator available without interrupting working historical access, so it always goes first when it applies.

Compare the failing client against a working one #

Two applications on the same workstation can address different servers, ports or identities, so compare the connection definitions themselves, not the display names the applications show. Record both definitions before changing anything.

Setting Where to check it What a mismatch causes
Server name or address each client's historical connection definition requests go to a different endpoint
Port the client definition against the server's configured listener port the client never reaches the intended service
Authentication mode the client's login settings the server evaluates a different identity than expected
Runtime identity the account the client process actually runs under Windows authentication presents the launching account, which may not be the expected operator
Database access the server-side login and its permissions a valid login still cannot read historical data
Client version and updates the installed-product information on each node a reported defect can apply to one build only

Both authentication modes work when the installation is configured for them. Microsoft documents Windows Authentication as SQL Server's default mode and the more secure of the two, which is a reason to standardize on it rather than to flip modes mid-fault. Switching modes changes which identity the server evaluates and can mask a rejected login or a missing permission instead of fixing it, so keep the site's approved method: verify the account actually running the client for Windows authentication, the configured login and its access for SQL authentication, and keep credentials out of screenshots and diagnostic records either way.

Test the server and the network path #

Work this branch when every client fails, or after a comparison shows both clients aimed at the same healthy definition.

  1. On the historian server, confirm the historian components and the SQL service backing the connection are running, and read their logs for startup failures or interruptions at the failure's timestamp.
  2. Read the TCP port of that SQL endpoint from the server's network configuration. 1433 is the default for a default SQL Server instance; a named instance can carry a different configured number, which the SQL Server Browser service answers on UDP 1434. Test the configured number, because a 1433 test against a server listening elsewhere proves nothing.
  3. From the failing workstation, run the port test worked through in the next section. Microsoft's connectivity documentation gives the PowerShell cmdlet Test-NetConnection as the quick network test for reaching a SQL Server instance, and tnc is the short alias for the same cmdlet.
  4. Run it twice: once with the server name the client uses, once with the server's verified IP address. Name fails while the IP succeeds: name resolution on that workstation. Both fail: routing, firewall policy or the listener. Both succeed with a still-failing client: authentication or the connection definition.

If the failing client runs as a virtual machine, add its virtual adapter to the path: a guest whose adapter mode changed can lose plant-network reachability while everything inside it stays up, and fixing a VM remote PLC connection with a NAT adapter walks that adapter-level fix.

Testing TCP 1433 from a failing trend client node #

The inputs: the failing node is a remote trend client workstation, the case where the server is healthy locally but a remote node cannot connect. The endpoint under test is the server name or address configured in that client's connection definition, written here as the placeholder HistorianNodeNameOrIPAddress, a string you replace with your value. The port is 1433, an integer, valid here because the server's network configuration shows 1433 as the historian's SQL port; any other configured number replaces it in the command below.

The rule in words: a trend client must complete a TCP connection to the server's SQL endpoint before it can log in or request a single tag, and Test-NetConnection opens exactly that connection, an integer port against a string name or address, and reports whether the handshake finished. That isolates the network layer from everything behind it. From the failing workstation:

tnc HistorianNodeNameOrIPAddress -port 1433

The arithmetic: the client's definition targets HistorianNodeNameOrIPAddress on port 1433, and the command tests that same name on port 1433. Configured 1433 = tested 1433, so a pass clears the network layer for exactly the endpoint the client uses, and for nothing else, because another port on the same server or the same port reached from another computer is untested. If the output echoes a remote port other than 1433, the command does not match the client's definition and the result says nothing about this fault.

The result: the decisive output is TcpTestSucceeded, a Boolean. True means the TCP handshake completed and a listener accepted the connection at that address and port; False means no connection opened at that endpoint.

The decision, by threshold: False on the name and False on the verified IP means fix the network path, which is routing, firewall policy or a listener not bound on that port, and no credential change helps while this stays False. False on the name but True on the IP means correct name resolution on that workstation and retest. True while the client still fails means the network layer is cleared for that endpoint, so stop testing the network and move to authentication and the connection comparison. True does not prove the listener is the intended SQL service, that the login is accepted, or that a historical query returns data; those are the next layers, not this test.

Sequence restarts and firewall changes through your site's change process: a production historian serves live trends, and an unplanned restart pulls them away from operators using them.

  1. Record the working and the failing connection definitions before changing anything, and keep one working historical display open as your reference.
  2. Restore the faulted layer in order: a stopped component first, then the network path, then the endpoint definition, then the rejected identity. Each layer hides the ones behind it, so an early fix can clear several symptoms at once.
  3. For a client-only failure, correct that client's connection to the verified server and the approved authentication configuration. Do not edit tag definitions, and leave the working clients untouched.
  4. Reconnect with the client's normal connection controls and repeat the same historical query after each change, so every result tells you which correction mattered.
  5. If endpoint and authentication checks pass and the failure persists, stop changing configuration and move to the version checks below.

With connections restored, moving history into another system is its own job: backfilling PI from an existing plant historian covers the transfer side.

When the basics pass: versions and reported defects #

Version-specific defects are the last branch, and they only matter once services, network and authentication all check out. Exact builds matter because a reported defect applies to named client and server versions, and a fix present in one build can be absent from another.

Two reported cases show the pattern. One plant needed a corrective update after a network hardware upgrade, described as a known but rare defect. A public report on Historian 2023 R2 described the INSQL OLE-DB provider failing after server restarts. Treat both as reported symptoms rather than verified causes for your installation: they say where to look, which is versions and the vendor's support portal, not what to change.

The vendor's own troubleshooting checklist starts the same way: search the support portal for the exact error text. Carry the client and server versions with update level, the configured endpoint and port, the authentication mode, the port-test result from the failing node, and timestamps that line up with the server logs.

Questions asked

Which account does Windows Authentication use for a trend client?

The one that launches the client process. A trend client started from a desktop shortcut presents the signed-in user; the same client launched as a service, from a scheduler or under saved credentials presents that identity instead. When a trend client works for an engineer but fails for operators, compare the two launch contexts and their accounts before changing anything on the server.

What does a connection test on the historian server itself prove?

A local success proves the historian and its SQL listener are up and accepting connections on that node, which clears the services branch. It does not clear the network branch: the remote client crosses switches, firewalls and VLAN rules the local test never touches. So treat it as half the answer, and always repeat the port test from the failing workstation before blaming credentials.

Can a virtual machine snapshot rollback cause a trend client connection failure?

Yes. Rolling a historian or client virtual machine back to an earlier snapshot can revive an old network configuration, a previous address, a stale adapter or a saved credential the server no longer recognizes, while the current setup looks right inside the guest. After a rollback, re-run the port test and recheck the client's connection definition against the live server before changing anything else.

Why does the historian answer every other node but one workstation cannot reach port 1433?

Because the fault sits on that workstation's path, not on the server. A local firewall rule, a stale route, an out-of-date DNS record or a changed VLAN assignment can block one machine while the rest keep working. That is why the port test belongs on the failing node: a result from any other computer tests a different path and proves nothing about the broken one.

Keep reading

  • Troubleshooting

    Why Power BI cannot connect to FactoryTalk Historian

    Power BI has no native FactoryTalk Historian connector because the PI-based archive is a proprietary time-series store, not SQL tables. Reach it through the licensed 64-bit OLE DB or ODBC client provider installed on the Power BI machine, aggregate on the server before import, or stage the data through an Excel PI DataLink workbook when no provider is licensed.

    Edge computing and IIoT

  • How-to

    Backfilling PI from an existing plant historian

    Split the historian-to-PI backfill into four independent stages — extract, intermediate store, external compression, destination load — and qualify throughput with a measured pilot. Rows divided by elapsed seconds, not tag count, sets the schedule: 12,000 tags from 2020 at 500 rows/s is about 203 days. Check the native Historian 2023 replication route before writing pipeline code.

    Edge computing and IIoT

  • How-to

    Deploying Edge Data Store in Docker on a controller PC

    Run Edge Data Store as a Docker container on the controller-class edge PC and let its OPC UA adapter ingest the local server over localhost. Define egress targets, buffering and forwarding schedules in the JSON configuration, size the buffer from tag count, scan rate and outage target—50 tags at 1 event/s across an 8-hour outage is about 288 MB—and sign off only after a live pulled-cable test.

    Edge computing and IIoT

SQL Server CAL count for historian clients: how to size it