Troubleshooting
Fix Ignition SQL Server connection faults after an upgrade
Short answer
After an Ignition 8.1.x upgrade, a faulted SQL Server connection is almost never a credentials problem: the new Java runtime disables TLS 1.0 and 1.1 (builds from 8u291), and a Microsoft JDBC driver as old as 4.0 cannot run on it. Read the Gateway exception, then patch SQL Server to TLS 1.2 per KB3135244 or replace the driver; the reported working pair was 9.4 in place of 4.0.
On this page
- Why did the SQL Server connection fault right after the Ignition upgrade?
- Where do I read the actual exception in the Gateway?
- Is the fault TLS, the driver, the certificate, or the network?
- How do I confirm SQL Server 2014 can negotiate TLS 1.2?
- Which JDBC driver version is the Gateway actually loading?
- Should I re-enable TLS 1.0 and 1.1 in Java instead of patching?
- How do I apply the fix and verify the connection?
- Questions asked
The fault sits in the layers the upgrade changed, not in your credentials. Moving Ignition from 8.1.5 to 8.1.12 replaced the Java runtime under the Gateway, and Java builds from 8u291 disable TLS 1.0 and 1.1 by default — a SQL Server 2014 instance that has not been updated for TLS 1.2 cannot finish the handshake. The second documented cause is the Microsoft JDBC driver: a driver as old as version 4.0 does not run on the newer runtime. Read the Gateway exception, classify it against the table below, and fix that one layer — patch the server's TLS support or replace the driver. Password resets and rebuilt connections change nothing when the failure happens before authentication.
Why did the SQL Server connection fault right after the Ignition upgrade? #
Because the upgrade changed the software under the connection, not the connection: Java builds from 8u291 disable TLS 1.0 and 1.1 by default, and a Microsoft SQL Server JDBC driver as old as 4.0 will not run on the newer runtime.
Two branches cover these faults, and both were introduced by the upgrade rather than by anything you typed into the connection.
- TLS negotiation. The newer Java runtime refuses TLS 1.0 and 1.1 by default. SQL Server 2014 gains TLS 1.2 only through the updates Microsoft documents in KB3135244, so an unpatched instance offers exactly the protocols the Gateway no longer accepts. The handshake ends before SQL Server ever reads a login packet.
- JDBC driver. In one reported installation, the move to 8.1.11 took Java to build 1.8, the loaded Microsoft SQL Server JDBC driver 4.0 did not support that runtime, and replacing it with 9.4 restored the connection. Treat the layer as the lesson and verify the pair on your own Gateway.
Both failures occur ahead of authentication. That is why resetting the database password, recreating the connection, and restarting the SQL service produce the identical fault, and why you should leave the existing connection in place while investigating — it is your comparison baseline.
Where do I read the actual exception in the Gateway? #
Open Configure → Database → Connections and click Details on the faulted connection; the panel shows the full error, and because the Gateway re-tests an unavailable connection every 10 seconds, the status you see is current.
The Gateway log carries the nested cause underneath the top-level connection failure; capture the whole chain, because the decisive line is usually two or three causes down. Keep credentials out of screenshots and support packages while you collect them.
The retest cycle also sizes the log noise you are about to read. With one retry every 10 seconds: 60 s ÷ 10 s = 6 attempts per minute, 6 × 60 = 360 per hour, and 360 × 24 = 8,640 per day, each attempt writing the same exception chain. Read one complete exception — the other 8,639 add volume, not information.
Is the fault TLS, the driver, the certificate, or the network? #
The exception text decides it: a protocol or handshake failure is the TLS layer, a driver-loading or unsupported-Java error is the JDBC layer, a certificate or server-name failure is identity validation, and a timeout or refused connection is the endpoint.
| What the exception says | Layer that failed | What to do |
|---|---|---|
| Protocol negotiation or handshake failure | Java and SQL Server share no enabled protocol | Patch SQL Server to TLS 1.2 per KB3135244 |
| Certificate path, trust, or server-name mismatch | TLS works; identity validation fails | Correct the chain, validity, or hostname |
| Driver load or unsupported Java version | JDBC driver incompatible with the runtime | Replace the driver through Ignition's driver management |
| Login failed for the configured user | Authentication | Credentials, authentication mode, database mapping |
| Timeout or connection refused | Endpoint reachability | DNS, listener, and firewall, tested from the Gateway host |
Test the endpoint from the Gateway host, not from an engineering laptop on a different network path. An open port proves something answers there; it proves nothing about which TLS protocol the instance will negotiate. When the row you matched is login or endpoint, the procedure for reconnecting an edge PC to SQL Server after connection loss covers the reachability and authentication checks in order.
How do I confirm SQL Server 2014 can negotiate TLS 1.2? #
Match the instance's exact build and patch level against Microsoft's KB3135244 TLS 1.2 support guidance — the label SQL Server 2014 alone does not answer it, because 2014 reaches TLS 1.2 through documented updates.
- Read the instance's full version and patch level, plus the Windows TLS configuration it runs on.
- Match that build against KB3135244, which lists what each SQL Server release needs for TLS 1.2.
- Read the Java version bound to the Gateway service — the upgrade replaced it, and a separately installed runtime may still answer elsewhere on the machine.
The arithmetic of the failure is the protocol overlap. After 8u291 the Gateway's enabled set excludes TLS 1.0 and 1.1. An unpatched 2014 instance that offers only those two protocols shares 0 of them with the client, so negotiation fails regardless of credentials, driver, or firewall state. The decision follows directly: update the server until TLS 1.2 is available on that instance, then re-test the connection.
Which JDBC driver version is the Gateway actually loading? #
The one the connection actually selects — verify it by reading the installed Microsoft SQL Server JDBC driver version and its Java compatibility, because a newer jar file on disk proves nothing about what the Gateway loads.
Identifying the MSSQL JDBC driver version in Ignition 8.1 is the first half of this check; the second half is that version's documented Java compatibility.
Replace the driver through Ignition's supported driver-management procedure, confirm the faulted connection selects the new version, and re-read the connection properties before you call it done: encryption and certificate-validation behavior differ between driver releases, and a jump as large as 4.0 to 9.4 crosses years of changed defaults. Take the driver from Microsoft or an authorized distribution channel, and keep the previous driver file plus the connection settings so the change reverses cleanly.
Should I re-enable TLS 1.0 and 1.1 in Java instead of patching? #
No — patch the server. Re-enabling TLS 1.0 and 1.1 in the Gateway's Java runtime weakens security for every connection that runtime makes, and rolling back Ignition only postpones the identical failure to the next upgrade.
Re-enabling deprecated protocols inside the Gateway lowers the security posture of every TLS session that runtime initiates — every database, every outbound connection — not just this SQL Server. If production pressure forces it as a bridge, make it a time-boxed change with approval, and remove it the day the server update lands. An Ignition rollback is the same category: a controlled recovery from a compatible Gateway backup that buys time for the real repair, which remains the KB3135244 update or the driver replacement.
How do I apply the fix and verify the connection? #
Change one layer at a time, re-test the same connection, and read the new exception — when the error moves from a handshake failure to a login failure, the transport is fixed and you are left with a smaller, ordinary problem.
- Back up the Gateway configuration and record the connection settings, loaded driver version, Java version, and SQL Server build.
- Apply the one fix the exception selected — the server-side TLS update or the driver replacement — and nothing else in the same window.
- Restart only what the updated component's documented procedure requires.
- Re-test the existing connection and read the new exception against the original.
- Confirm a real write: a known tag, value, and timestamp present in SQL Server, not just a Connected status panel.
- If connections are healthy but history queries still fail, the remaining fault is in the query layer — tracing a SQL log error after a historian migration walks that read.
Questions asked
Did we lose tag history data while the connection was faulted?
Nothing reached SQL Server during the fault; a faulted connection executes no queries. Ignition buffers database writes in its store-and-forward pipeline while a connection is unavailable, so after recovery check that pipeline's status and let it drain before deciding anything was lost. How much it holds depends on how it was configured.
Is rolling Ignition back to 8.1.5 an acceptable fix?
As a recovery action under change control, yes; as the repair, no. The rollback restores the older Java runtime and the connection comes back, but the TLS and driver incompatibility is still there and returns with the next upgrade attempt. If you roll back from a compatible Gateway backup, schedule the SQL Server update or driver replacement immediately after.
Where should the replacement JDBC driver come from?
Microsoft or an authorized distribution channel, and only the version you validated against your Ignition release and Java runtime. Third-party jar repositories are not a source for software that holds your database credentials. Keep the previous driver file and the connection settings so the change reverses cleanly if a connection property behaves differently.
Why does the Gateway log repeat the same exception all day?
Because it keeps retrying: the documented 10-second retest is 6 attempts a minute and 8,640 a day, and each writes the full chain. The repeats carry no new information. Capture one complete exception with its nested cause, and filter the repeats when you package logs for support.
Do PostgreSQL or MySQL connections break the same way after an upgrade?
Yes, at the same layer. A documented PostgreSQL case describes connections failing after the database server was upgraded because the JDBC driver Ignition had loaded was older than the server accepted. Ignition ships drivers for the major databases, and the remedy is identical: inventory the loaded driver and move it to a version compatible with both the new server and the running Java.