Hello,

I'm managing a website with multiple nodes, and I noticed that sometimes information isn't well shared between the nodes. I identified that the nodes read data properly from the database, but the problem is sometimes information isn't sent to the database.

Say for example, a user is connection on node a
I connect as an admin on node b, and I deactivate the user on node a

He still has access to the site as if nothing happened.
I check the database, his active status is still set to true.
I use the search in liferay's Users portlet in the backend and it doesn't return the user when selecting "active".
It returns the user when selecting "Inactive", but when I click on "Actions" on this user, It allows me to deactivate the user once again.
I suppose SOLR updated the user's status properly, but liferay didn't push the new value to the database.

How is it possible that some information isn't sent straight to the database?
It used to work fine but recently we've been having a lot of problems related to this.

Any help is appreciated
Thank you very much
David

Sounds like you have dirty caches. Have you clustered the cache layer and tested that it infact works?

Hello Mika, thanks for replying

I'm pretty sure this was done way back when we started the project to split the site into multiple nodes. Though perhaps the configs changed at some point or a file was overwritten. I was not part of the project at that point and I have no experience in this.

Would you be able to provide me some information on how to cluster the cache layer properly? Are there any configs that we need to make sure are set for this to work properly?
I found http://www.liferay.com/documentation/liferay-portal/6.0/administration/-/ai/distributed-cachi-4 , would you say it is related to my problem? Though we recently checked those settings and we didn't find anything alarming... Perhaps we missed something?

Thank you very much!
David

I've also checked http://www.liferay.com/community/wiki/-/wiki/Main/Portal+Properties+6.0.5#section-Portal+Properties+6.0.5-Ehcache and our properties are similar, and use multi-vm-clustered... Is it possible that some configs in the xml files themselves are not set properly?

I notice it says to edit the file under ehcache.multi-vm.config.location to enable clustered caching... I've asked for the xml files in prod I will come back with some more information soon!

One easy way to test if it's working is to open a page on separate browsers on each of the nodes. Then do a change like drop new web content portlet on that page and reload all the other browser to see if the change is immediately visible on the other node. If it's not then your cache cluster is not working. You should also try changing content on web content article to see that change is propagated to all nodes.