Summary

In Confluence 6.x and later, you can't create new pages or edit existing ones due to the editor not loading. The logs show an error similar to this: 'Could not initialize class org.xerial.snappy.Snappy'.

Check the mount point

The root cause of this issue lies in the operating system's mount point(s), whereby the /tmp partition has been mounted as noexec. Mounting a file system as noexec means it does not permit the execution of executable binaries in the mounted file system.

Resolution 1

The /tmp partition needs to be remounted as exec. Contact your Linux administrator so they can perform this fix. You can also use this command to remount the partition and resolve the issue:

mount -o remount,exec /tmp

Or, if you don't allow /tmp to be mounted with exec permissions, you can add the following argument to the <confluence-home>/synchrony-args.properties file:

java.io.tmpdir=/path/to/your/own/temp/directory

Please check the Configuring Synchrony for Data Center guide on how to use the synchrony-args.properties file depending on your version. The way to pass JVM arguments to managed Synchrony nodes has changed in versions 7.20 and newer and differs if you use standalone mode.

Also, make sure that /path/to/your/own/temp/directory is owned by the operating system user account that runs Confluence.

Resolution 2

Alpine Linux distribution by default won't have libc6-compat package installed, which provides the symlinks for libc.so.6 / libm.so.6. Check if the package is installed, and if not, install it.

apk add libc6-compat

Resolution 3 (when this affects the Jira Issues Macro and after checking Resolution 1)

1. Shutdown Confluence
2. Clear the plugin cache by removing the following directories:
   
   • <confluence-home>/bundled-plugins
   • <confluence-home>/plugins-cache
   • <confluence-home>/plugins-osgi-cache
   • <confluence-home>/plugins-temp
   • <confluence-home>/bundled-plugins_language


     This will remove the plugin cache files only and will not permanently remove any plugins. These directories will be recreated during the next Confluence startup. However, some third-party plugins may store metadata in these directories, so we recommend creating a temporary backup of the directories.

3. Start Confluence.

Resolution 4

1. Shutdown Confluence
2. Add the following system property:
   

   --add-opens=java.base/java.nio=ALL-UNNAMED
   --add-opens=java.base/sun.nio.ch=ALL-UNNAMED
   --add-opens=java.base/sun.nio.cs=ALL-UNNAMED

3. Start Confluence

Notes

Issues with migration from Linux to Windows

This problem may also occur when migrating from Linux to Windows. After the first startup on the new server, you can see the following error in the atlassian-synchrony.log file:

{"synchrony":{"entity":"/Customer-Instance-afa0320d-c0c7-3fcf-973e-38159f4035d0/confluence-17171501","ns":"synchrony.sync.hub","level":"warn","throwable":"java.lang.UnsatisfiedLinkError: /tmp/snappy-1.1.0.1-cc895952-7b1a-4dad-aa7d-c51ee5bf9595-libsnappyjava.so: /tmp/snappy-1.1.0.1-cc895952-7b1a-4dad-aa7d-c51ee5bf9595-libsnappyjava.so: failed to map segment from shared object: Operation not permitted
...
{"synchrony":{"message":"synchrony.http.entity-api [warn] Error in put-entity","entity":"/Synchrony-ec07ec53-37cb-3637-be42-10a55a8b6b18/confluence-3736095","ns":"synchrony.http.entity-api","level":"warn","throwable":"clojure.lang.ExceptionInfo: Could not initialize class org.xerial.snappy.Snappy

As we can see, Synchrony is still looking into the /tmp directory for the snappy file. To refresh the directory, to pass a new valid path to the synchrony-args.properties file as in the following example and restart Synchrony:

java.io.tmpdir=C:\\Progra~1\\Atlassian\\Confluence\\temp

Issues with collaborative editing

A bug occurs when using Java 17 and the /tmp directory is mounted with noexec. This causes the same error in the logs because the library can't automatically use a different tempdir.The bug ticket is tracked here: CONFSERVER-98444 - Collaborative editing is not working when Java 17 is used and /tmp is noexec

Support of ARM

You'll need to use x86 hardware or 64bit derivatives of x86 hardware as mentioned in Supported Platforms. We have a feature request to support ARM: CONFSERVER-60187 - Support ARM platforms

Timeout when rendering content

Problems involving the 'Could not initialize class org.xerial.snappy.Snappy' error can occur within two different contexts.

First, it might occur while trying to create a new page or edit an existing page.  In this case, the UI displays a message that it's taking a long time to load the editor. As this message is somewhat generic and can occur as a result of different errors, we'll need to take a closer look into the logs to check for occurrences of the org.xerial.snappy.Snappy error. Second, it's possible for this error to occur while trying to render content within an instance of the Jira Issues macro.  In this case, the macro content will include the message jiraissues.unexpected.error.

Diagnosis

Environment

• Confluence 6.x and later
• Confluence is hosted in any of the following operating systems:
  • Debian
  • CentOS
  • Ubuntu
  • Red Hat Enterprise Linux
  • Alpine Linux

Diagnostic Steps

To further diagnose this issue, we'll need to examine the application and Synchrony logs.

Page Creation/Editing

In cases where this is related to a page editing issue, the atlassian-synchrony.log files will mention some variation of org.xerial.snappy.Snappy.  Here are some examples:

:cause "Could not initialize class org.xerial.snappy.Snappy"

java.lang.NoClassDefFoundError: Could not initialize class org.xerial.snappy.Snappy

java.lang.UnsatisfiedLinkError: /tmp/snappy-1.1.0.1-f08e07cc-9064-42a4-9af9-b062417b146f-libsnappyjava.so: /tmp/snappy-1.1.0.1-f08e07cc-9064-42a4-9af9-b062417b146f-libsnappyjava.so: failed to map segment from shared object

clojure.lang.ExceptionInfo: Could not initialize class org.xerial.snappy.Snappy {:type :server-error, :source :server}

Exception org.xerial.snappy.SnappyError: pure-java snappy requires access to java.nio.Buffer raw address field

Jira Issues Macro

In cases where this is related to a Jira Issues macro instance, we'll see messaging like the following within the atlassian-confluence.log files:

2020-12-10 19:55:40,510 ERROR [StreamsCompletionService::thread-10] [xhtml.view.macro.ViewMacroMarshaller] handleMacroExecutionException Error rendering macro: jira
 -- url: /plugins/servlet/streams | traceId: a3968fc5d8e9894b | userName: charlie
java.lang.NoClassDefFoundError: Could not initialize class org.xerial.snappy.Snappy
	at com.atlassian.confluence.extra.jira.JiraIssuesManager$ByteStreamBasedSupplier.compress(JiraIssuesManager.java:117)
	at com.atlassian.confluence.extra.jira.JiraIssuesManager$ByteStreamBasedSupplier.<init>(JiraIssuesManager.java:107)

Additional Diagnosis

Looking further, if the /tmp partition is mounted on the server, we see that it's mounted as noexec. We can check this by running the following command on the server's terminal:

mount | grep /tmp

You should see results similar to the following, which confirms that /tmp has been mounted as noexec:

/dev/sda7 on /tmp type ext4 (rw,noexec,nosuid)