IMD SolutionComputing LibraryShop main store

Troubleshooting

Ignition Edge stuck starting after a Java trust store corruption

Edge computing and IIoT Published 6 min read

Short answer

The gateway page stays at Gateway is starting because module initialization dies when Java cannot build its default SSL context — wrapper.log shows java.security.NoSuchAlgorithmException for DefaultSSLContext under SunJSSE. Copy cacerts and cacerts.bak out of lib\runtime\jre-win\lib\security in the installation the service runs, rename cacerts.bak to cacerts, restart the service, and the gateway comes up.

On this page
  1. What the "Gateway is starting" page proves
  2. Follow the exception chain, not the warnings
  3. Diagnosing the Edge 8.1.3 startup hang from the 11:27:30 log lines
  4. Restore the trust store in the installation the service runs
  5. Why an unclean shutdown does this, and what prevents a repeat
  6. Questions asked

The gateway is not frozen; one startup thread has died while the web server keeps serving. When the Ignition Edge page never leaves "Gateway is starting" after an unclean Windows restart, wrapper.log ends the guesswork: a java.security.NoSuchAlgorithmException naming DefaultSSLContext under provider SunJSSE means the bundled Java runtime's certificate trust store is damaged, and module and licensing initialization cannot build the default SSL context. The repair is one file. Copy cacerts and cacerts.bak out of lib\runtime\jre-win\lib\security in the installation the service actually runs, rename cacerts.bak to cacerts, and restart the service. On the documented Edge 8.1.3 gateway this restored startup on the next service start; no reinstall, no network changes.

What the "Gateway is starting" page proves #

The page loading at all tells you the failure sits inside gateway initialization, not in front of it. The browser request reaches a running web component; what never completes is the module manager behind it. Three lines from the failing attempt in wrapper.log separate the two:

wrapper.log observation What it establishes Decision it forces
HTTP connector started on 0.0.0.0:8088 The web listener bound to every IPv4 interface The page loading is expected; keep reading
SSL connector started on 0.0.0.0:8060 The inbound SSL connector initialized Inbound TLS is up; the fault is elsewhere
internal database "config.idb" started up successfully The configuration database is intact Do not rebuild the internal database

The decision from those three lines: leave the network, the address, and the internal database alone, and work inward to the exception that follows them.

Follow the exception chain, not the warnings #

Immediately after those lines, the execution engine reports a one-shot startup task throwing an uncaught exception:

java.lang.InternalError: java.security.NoSuchAlgorithmException: Error constructing implementation (algorithm: Default, provider: SunJSSE, class: sun.security.ssl.SSLContextImpl$DefaultSSLContext)

Read to the deepest Caused by and the stack under it rather than stopping at the first warning. The stack runs from HttpClientBuilderImpl.build through LicenseManagerImpl.createLeasedActivationStrategy to ModuleManagerImpl.init: while modules initialize, the licensing manager constructs an HTTP client, and constructing the default SSL context loads the runtime's trust material before any bytes reach a network. An unreachable licensing server would surface as a connection failure against a remote host; here nothing remote appears in the trace at all. The exception's name is misleading in this position — it reports that the default SSL implementation could not be constructed, not that an algorithm is missing from the platform.

The Illegal reflective access warnings about IgnitionSocketImplFactory a few lines earlier are long-standing compatibility warnings; they preceded months of healthy startups and are not the fault. Save the complete exception text, including every Caused by, before changing anything.

Diagnosing the Edge 8.1.3 startup hang from the 11:27:30 log lines #

Inputs, each from the deployment record or the log itself:

  • Host and history: Windows 10 PC; the Edge gateway and its Vision project had run for about two months.
  • Software: Ignition Edge 8.1.3, build b2021030309, from the attempt's first line, "Starting Ignition 8.1.3 (b2021030309)".
  • Trigger: one failed boot that ended on the Windows recovery screen; every restart after that was clean, and the gateway never started again.
  • Log window: gateway start line at 11:27:29; HTTP connector on 0.0.0.0:8088, SSL connector on 0.0.0.0:8060, config.idb opened, and the uncaught exception all at 11:27:30; Jetty logged Started @6717ms.
  • Size reference for judging the store: a separately documented JDK case in which a corrupt Windows cacerts measured 160 kB against 252 kB for the known-good store of the same release.

The rule in words: the last healthy subsystem before the exception, plus the deepest Caused by, names the failed dependency; the time between them bounds how much log you must read; and a trust store is judged corrupt by comparison against its shipped backup, not by eye.

The arithmetic:

  • Fault window: 11:27:30 − 11:27:29 = 1 s. The entire failure occupies one second of log, in the same second the connectors came up.
  • On Jetty's internal clock: 6717 ms ≈ 6.7 s from server start to connectors up; the exception lands inside that same second.
  • Corruption signature: 252 kB − 160 kB = 92 kB missing, and 160 kB ÷ 252 kB ≈ 0.63 — the corrupt store was about 63% of the intact size.

Result: the deepest frame is java.security.Provider$Service.newInstance constructing the default SSL context, with no socket and no remote host anywhere in the trace. The failed dependency is the local trust store.

Decision: with cacerts suspect, apply the size rule first. A cacerts far smaller than its cacerts.bak, or near zero bytes, is corrupt — do the swap below. Two files of equal size point to a read-permissions problem under the service account instead, and a rename will not fix that.

Restore the trust store in the installation the service runs #

One trap first. The failing service ran from C:\Users\HMI\Desktop\Ignition-Edge-windows-64-8.1.3 — the jar and webapp paths in wrapper.log say so — while the same procedure is often written against C:\Program Files\Inductive Automation\Ignition. Both installations can exist on one PC, and editing the wrong one repairs nothing. Confirm before you copy:

  1. Identify the active installation: read the jar or webapp paths at the top of wrapper.log, or the executable path on the gateway service's Properties page in the Windows Services console.
  2. Stop the gateway — the Services console or stop-ignition.bat from that installation — and confirm Windows reports it stopped. Schedule the stop around the running project.
  3. In <installation>\lib\runtime\jre-win\lib\security, copy both cacerts and cacerts.bak to a folder outside the runtime, and record both file sizes.
  4. Delete cacerts, then rename cacerts.bak to cacerts.
  5. Start the gateway with start-ignition.bat or the Services console.
  6. Verify in wrapper.log that startup passes module and licensing initialization with no new NoSuchAlgorithmException, and that the web page leaves "Gateway is starting".

If certificates from an internal CA were imported into cacerts after installation, re-import them into the restored store — the backup predates them. With the gateway back, check the rest of the shutdown's aftermath: if the Edge historian disappears after the next restart instead, that is a separate fault, and why the Ignition Edge historian keeps disappearing covers it.

Why an unclean shutdown does this, and what prevents a repeat #

cacerts is an ordinary file on the host disk, and the recovery-screen boot cycle is what damaged it in the documented case. Java reads it only when the default SSL context is constructed, which is why the gateway ran for months and then failed on the first startup after the corruption — the damage sat silent until something built an SSL context.

To keep it from recurring:

  • Put the edge PC on a UPS so Windows shuts down and boots cleanly instead of cycling through recovery.
  • Set the gateway service's Windows recovery options to restart it after a failure, so a startup fault does not leave the page hanging unattended.
  • Keep a current gateway configuration backup, so a runtime repair never puts the project at risk.

If the same exception returns after a verified swap in the verified installation, and the service account can read the new file, reinstall the gateway over the existing installation with the configuration backed up first. Reinstall is the fallback, not the first move; in the documented case the file swap alone restored service.

Questions asked

Why does the gateway web page still load if startup failed?

Because the web server starts before module initialization. The lines binding 0.0.0.0:8088 and 0.0.0.0:8060 mean the listeners attached to every IPv4 interface on the PC — 0.0.0.0 is a binding address, not a destination, so browse to the PC's actual hostname or IP address. What never finishes is the module and licensing thread behind the page, which is why the page loads and startup still hangs.

Is it safe to delete cacerts without keeping a copy?

No. Copy both files to a folder outside the runtime before deleting anything. The copies preserve rollback if the swap does not cure the fault, and the original file is your evidence: a truncated or zero-length cacerts confirms corruption, while an intact-looking file that will not read points to a permissions problem. The repair is only reversible while you keep both files.

The fix worked, but the fault returned after another power cut. What now?

Fix the power first. Put the PC on a UPS so Windows shuts down cleanly, and set the gateway service's Windows recovery options to restart it automatically. Then compare the cacerts and cacerts.bak sizes after each event — if the store is damaged repeatedly on one machine, the drive or its filesystem is the suspect, and that is the repair to make.

Did the shutdown damage the Vision project or its tag data?

No. The fault fires during module and licensing initialization, before the project loads, and the internal database config.idb opened successfully in the failing attempt — the project and its configuration survived. Verify after the swap that the project loads and its tag provider reconnects. Only a careless reinstall would put that data at risk, so take the configuration backup before any reinstall.

Will restoring cacerts.bak drop certificates we added later?

It can. cacerts.bak is the backup shipped with the runtime, so any CA certificate imported after installation is missing from it. Re-import those certificates with the bundled Java keytool once the gateway is up, or outbound TLS to systems whose certificates chain to your internal CA will start failing validation — a different fault that looks like a network problem.

Keep reading

  • Troubleshooting

    Historian stores zero after an MQTT edge node reconnects

    A 0 stored just before the true value at every MQTT edge reconnect is a reported pattern on Ignition 8.1.5 through 8.1.15 with the MQTT Transmission and Engine modules 4.0.6 through 4.0.10. Confirm it by querying stored records around each reconnect, exclude those zeros from process data, and remove them with a timestamped rule instead of a blanket zero filter.

    Edge computing and IIoT

  • Troubleshooting

    PI-to-PI interface service will not start from FactoryTalk Historian

    When the PI-to-PI interface service will not start, the fault sits in the local launch, access to the two historians, or the FactoryTalk Historian license, which must permit native OSIsoft clients. Capture the exact startup error, work the three gates in fixed order, then prove the destination with values, timestamps and quality.

    Edge computing and IIoT

  • Explainer

    Where the Ignition Edge tag historian writes its data

    Ignition Edge has no configurable path for its local tag historian: the store and its store-and-forward buffer both write to the volume Ignition is installed on. For a 7-day history on fewer than 100 tags, add 3 to 5 GB of headroom to the main partition; an external SQL database connection is the only supported way to choose the volume.

    Edge computing and IIoT