What causes "Unable to obtain input stream from [{}]" in hibernate/hibernate-orm?

Verify the resource named in the message exists and is readable at bootstrap time (open it manually from the same classloader).

hibernate/hibernate-orm · error · PersistenceException

Unable to obtain input stream from [{}]

Error message

Unable to obtain input stream from [{}]

What it means

Thrown by PersistenceXmlParser.loadUrlWithJaxb() as a PersistenceException when URL.openConnection() succeeded but connection.getInputStream() raised an IOException while reading a persistence.xml resource. The resource URL resolved, but the actual stream could not be opened - the connection to the resource broke at read time. The resource name is included in the message.

Source

Thrown at hibernate-core/src/main/java/org/hibernate/jpa/boot/spi/PersistenceXmlParser.java:326

			}
			throw new PersistenceException( "Unknown persistence unit transaction type : " + value );
		}
	}

	private JaxbPersistenceImpl loadUrlWithJaxb(URL xmlUrl) {
		final String resourceName = xmlUrl.toExternalForm();
		try {
			var connection = xmlUrl.openConnection();
			// avoid JAR locking on Windows and Tomcat
			connection.setUseCaches( false );
			try ( var inputStream = connection.getInputStream() ) {
				return new ConfigurationBinder( classLoaderService )
						.bind( new StreamSource( inputStream ),
								new Origin( SourceType.URL, resourceName ) )
						.getRoot();
			}
			catch (IOException e) {
				throw new PersistenceException( "Unable to obtain input stream from [" + resourceName + "]", e );
			}
		}
		catch (IOException e) {
			throw new PersistenceException( "Unable to access [" + resourceName + "]", e );
		}
	}

}

View on GitHub (pinned to fad1729dce)

Solutions

Verify the resource named in the message exists and is readable at bootstrap time (open it manually from the same classloader).
Rebuild/redeploy the artifact to eliminate a corrupted or truncated jar.
On Windows/Tomcat, ensure no process locks the jar (setUseCaches(false) is already applied by Hibernate) and stop parallel redeployments during boot.
If the file lives on a remote URL, check network availability and server logs for the refused read.

Example fix

// before: jar replaced while app boots
// after: deploy atomically - build the war fully, then swap it in one step
// e.g. cp app.war.new tmp/ && mv tmp/app.war.new webapps/app.war

Defensive patterns

Strategy: try-catch

Validate before calling

// pre-open the descriptor from the same classloader before boot
URL url = Thread.currentThread().getContextClassLoader().getResource("META-INF/persistence.xml");
if (url == null) throw new IllegalStateException("persistence.xml not found");
try (InputStream in = url.openConnection().getInputStream()) { in.read(); } // fail with your own error

Try / catch

catch (PersistenceException e) {
    if (e.getMessage() != null && e.getMessage().contains("Unable to obtain input stream")) {
        log.error("persistence.xml unreadable (locked/removed/corrupt jar): {}", e.getMessage());
    }
    throw e;
}

Prevention

Deploy archives atomically: build the full artifact, then swap it in one step.
Keep classpath entries immutable while the application boots; no parallel repackaging.

When it happens

Trigger: Opening the persistence.xml URL during parsing where the jar/file was deleted, swapped, or locked between URL resolution and stream read; a jar entry that reports a size but cannot be read (corrupt zip); permission denied on the file; connection refused for an http-hosted descriptor.

Common situations: Hot redeploy/repackaging on Windows where the old jar is locked or removed mid-boot; exploded deployments whose META-INF/persistence.xml permissions changed; partially uploaded or corrupted jar files; Tomcat/JBoss wars being modified while Hibernate scans them.

Related errors

AI-assisted analysis of hibernate/hibernate-orm@fad1729dce (2026-08-22). Data as JSON: /api/errors/abf5fe22a1b3d316. Report an issue: GitHub.