Hi!
We’ve been experiencing what looks like data corruption on two machines that we know of. It could be specific to those machines, but I’d like to compare notes with other people anyway.
It seems we get HiveMQ to crash reliably, and I noted the following in the log:
internal running controller killed restart handler
If I delete the data folder, HiveMQ can start up again.
The condition has happened at least once on a single install site, and at least twice on another. That could just be some machine-local issue such as malfunctioning memory or file system, so I’ve asked to install on a different machine. But I’d also like to start digging in case it’s not.
Have you encountered something similar?
I’m running HiveMQ 2025.4on Amazon Corretto JDK 21.0.8_9. I’m running it on Windows (unsupported, I know) as a daemon/service using the YAJSW Windows Service wrapper.
The reported error and the fact that deleting the data folder resolves it suggests a persistence inconsistency (typically caused by an unclean shutdown, abrupt JVM termination, filesystem issues, or external interference). On Windows (which is unsupported for production) and especially when using a wrapper like YAJSW, forced restarts or service timeouts can interrupt HiveMQ’s embedded persistence layer and leave it in a non-recoverable state at next startup.
I’d recommend:
- check the full log before that message for the real root cause (often an earlier persistence or I/O exception)
- verify the JVM is not being killed externally (service timeout, OOM, Windows Service Control Manager)
- exclude the HiveMQ directory from antivirus/real-time scanning
- ensure graceful shutdown (
bin/run.bat stop equivalent via wrapper)
If it recurs, please share the complete startup log from the failing attempt, as well as any relevant system logs.
Kind regards,
Finn