If activation fails due to a mismatch of the keys on the two instances, an error "Handshake information for activation was incorrect" is displayed. You can also find the error message in the logs:
To resolve the issue, generate a new public key and copy it to the public instance.
See Monitoring for more about logging and debugging Magnolia.
Non-EE subscriber detected
If you activate SSL certificates on your magnolia public instance, and configure everything so that any http request is redirected to https by default but leave the subscriber on the author instance configured to point to
http:// , any activation request will fail with the following message:
URL property in the configuration at the
/server/activation/subscribers/<subscriber-name>@URL path. Make sure it starts with
Simple exchange transport failed
Whenever Java attempts to connect to another application over SSL (e.g. HTTPS, IMAPS or LDAPS), it will only be able to connect to that application if it can trust it. The way trust is handled in the Java world is that you have a keystore, typically
JAVA_HOME/lib/security/cacerts , also known as the truststore. The error message in this type of failure will look like as follows:
To fix this, make sure that
Cleaning the errors
During the activation process Magnolia makes a copy of the current content on the public instance and stores it in the
mgnlSystem workspace. The system rolls back any changes using the backup copy if transactional activation fails for some reason. If the activation succeeds, the system overwrites the current content with the activated content and removes the backup copy.
Below tips and tools could help you either detect the issue or clean up unnecessary content before publishing:
The steps of an activation on Public can be viewed by turning the
DEBUG using the Log Tools app. Here we see an example of the steps during page activation.
Activation of page /travel/about/company
It is possible that something goes wrong while Magnolia is activating the content. For example, the server process is killed or the server loses power. This means that the backup copy may have gotten left in the
mgnlSystem workspace. Never clean up properly after the activation during the commit step.
Checking at startup
The system will automatically check for nodes. If a node is found in the
mgnlSystem workspace during startup this message will be recorded in the public log.
This is a sign that the server went down during the activation step. The backup copy remains indicating that the commit step was never reached. This means that the page may have not been activated correctly.
After a successful activation a cleanup procedure is run to remove the backup copies. This is performed during the commit step. If for some reason it cannot cleanup then a message like this will occur in the log.
To try and resolve the issue, clean the
mgnlSystem workspace yourself on the public instance. There is a cleaning tool in the Activation Tools app. This is the same procedure that is run during startup but triggered manually using this app.
After running be sure to check the log. If you seethen you need to take other actions. For example, restarting the server to release any locks that may still be present. Restarting will automatically call the cleanup procedure described above. If locks was the issue then you should see the system successfully remove the backup copies during the restart. However, if you still see repository exceptions in your log during startup then you most likely have a repository corruption issue. In that case probably the best course of action would be to recreate the workspace from scratch.
The JCR Browser app will allow you to visually inspect the workspace as well.
mgnlSystem workspace should normally be empty if there is no activation in progress. After cleaning, make sure the website tree is OK on your public instance.