Shared data folder for cluster

[William Fleming, módosítva 13 év-val korábban]

What is the best way to share the data folder between 2 clustered nodes. I have looked storing everything in database but I don't really want to go down this route as it would cause the mySQL server to explode in size. Then backing up the dumps would be a bigger hassle as the files would be huge. Right now I have NFS share mounted on both servers what other options are there?

Thanks
William

[Amos Fong, módosítva 13 év-val korábban]

Hi William,

This blog gives a decent overview http://www.liferay.com/web/james.min/blog/-/blogs/why-load-balancing-!%3D-clustering-in-liferay

[Hitoshi Ozawa, módosítva 13 év-val korábban]

As Amos replies, there are mainly 2 things to shares - 1. indexes 2. libraries.
I recommend using Solr for the indexes. From your question, I think you were mainly concerned about the libraries. Followings are standard options available. If you have MS Office documents, CMIS tool such as Alfresco can be used by using CMISHook.

##
## Document Library Portlet
##

#
# Set the name of a class that implements
# com.liferay.documentlibrary.util.Hook. The
# document library server will use this persist documents.
#
#dl.hook.impl=com.liferay.documentlibrary.util.AdvancedFileSystemHook
#dl.hook.impl=com.liferay.documentlibrary.util.CMISHook
dl.hook.impl=com.liferay.documentlibrary.util.FileSystemHook
#dl.hook.impl=com.liferay.documentlibrary.util.JCRHook
#dl.hook.impl=com.liferay.documentlibrary.util.S3Hook

[Juan Gonzalez P, módosítva 13 év-val korábban]

Hitoshi Ozawa:

As Amos replies, there are mainly 2 things to shares - 1. indexes 2. libraries.
I recommend using Solr for the indexes. From your question, I think you were mainly concerned about the libraries. Followings are standard options available. If you have MS Office documents, CMIS tool such as Alfresco can be used by using CMISHook.

##
## Document Library Portlet
##

#
# Set the name of a class that implements
# com.liferay.documentlibrary.util.Hook. The
# document library server will use this persist documents.
#
#dl.hook.impl=com.liferay.documentlibrary.util.AdvancedFileSystemHook
#dl.hook.impl=com.liferay.documentlibrary.util.CMISHook
dl.hook.impl=com.liferay.documentlibrary.util.FileSystemHook
#dl.hook.impl=com.liferay.documentlibrary.util.JCRHook
#dl.hook.impl=com.liferay.documentlibrary.util.S3Hook

Hi Amos,

I am using succesfully using cluster nodes local indexes replicated through cluster (alternative to using Solr). This is activated when setting cluster link and replicate indexes in portal-ext.properties (sorry I don't remember their specific name right now).

And AFAIK the best dl.hook.impl for cluster is AdvancedFileSystemHook (I've read it somewhere in docs/forum).

[Hitoshi Ozawa, módosítva 13 év-val korábban]

I am using succesfully using cluster nodes local indexes replicated through cluster (alternative to using Solr). This is activated when setting cluster link and replicate indexes in portal-ext.properties (sorry I don't remember their specific name right now).

lucene.replicate.write=true
cluster.link.enabled=true

We also tried this settings with the JBoss cluster and it was sometimes erroring due to duplicated index when post volume were high or when admin did a index recreate from on of the server.

[Juan Gonzalez P, módosítva 13 év-val korábban]

Hitoshi Ozawa:

I am using succesfully using cluster nodes local indexes replicated through cluster (alternative to using Solr). This is activated when setting cluster link and replicate indexes in portal-ext.properties (sorry I don't remember their specific name right now).

lucene.replicate.write=true
cluster.link.enabled=true

We also tried this settings with the JBoss cluster and it was sometimes erroring due to duplicated index when post volume were high or when admin did a index recreate from on of the server.

Yep I suppose this config depends on how many index operations are happening per minute, and network performance.