Error in balance with the environment not replicated web content

[Manuel Hoyos, geändert vor 7 Jahren.]

Hi.

We have a serious problem in balancing configuration Liferay 6.1 GA2. We have set up an F5 rolling with two web front, and two instances of LIFERAY.

Everything seems to have gone well, however when editing a web content on one node, the other node is not updated. When viewing the page containing the web content viewer, we note that this refers to the previous version, and not the latter created from the other node.

We activate the options:
- Lucene.replicate.write = true
- Cluster.link.enabled = true

And the content of forums, blogs, documents, etc., does sync smoothly.

I know it can not be happening, but we're pretty desperate.

In our test environment, where it works, the two nodes are identical copies, while in the production environment, they were not 100%, although the two nodes share database.

Any help is welcome.

Thank you

[David H Nebinger, geändert vor 7 Jahren.]

What kind of network is it? Have you tested jgroups to show that multicast messaging between the nodes is enabled? Are there any messages in the logs about forming into a cluster?

[Manuel Hoyos, geändert vor 7 Jahren.]

Thank you very much.

In our installation it is not enabled any distributed cache mechanism, or replication session.

Tomcat have configured servers, to perform rolling with mod_proxy_http.

And as I indicated, all content is replicated on both nodes as content, forums, blogs, alarms, etc. is changed, however this is not produced in the web content.

Where I can check messages that are exchanged in the catalina.out?

Certainly we are talking about the community version

Thanks again.

[David H Nebinger, geändert vor 7 Jahren.]

In Liferay, nodes form a cluster by sending multicast (broadcast) messages on the network and forms a cluster from all nodes that respond.

In some virtual environments and cloud environments, broadcast messaging is disabled so the nodes cannot form a cluster. Depending upon your server configurations, firewalls may also be preventing broadcast messages so the nodes cannot form a cluster.

This has nothing to do with session management (sticky or replicated), it is just in the ability of the nodes to form a proper cluster.