cancel
Showing results for 
Search instead for 
Did you mean: 

Alfresco 4.2.X CE with Load Balancing

alfresconewbie
Champ in-the-making
Champ in-the-making
Hi,

Hoping someone could clear something up for me;

We're looking at the Aflresco 4.2.X Community Edition, I understand the Hazelcast clustering code has been removed, however, what has been mentioned as an option is to run multiple application servers as standalone servers sharing a repository and db, fronting them with a load balancer which can handle sticky sessions.

Can anyone tell me what functionality we are losing by doing this? will this even work or will we end up with sharing issues at a process or document level?

Our architecture is based on:
- Alfresco CE running on two Ubuntu servers fronted with a Kemp Load balancer configured with stick sessions and affinity
-The repositiry is a Gluster FS share presented to both servers
-The DB is Postgres 9.2 fronted with PGPOOLII

Thanks
AN
4 REPLIES 4

mrogers
Star Contributor
Star Contributor
The Hazelcast clustering code has never been in community.  So it has not been removed.

Perhaps deleted content will still be shown.    Perhaps newly added content won't be shown.  Perhaps locked content will be shown as unlocked.   There will also be a collection of less obvious problems.

The bottom line is that community does not cluster.   

Hi,

Thanks for your response, albeit a little vague, could you clarify further? is data being persisted/cached at the application layer relevant to whether content has been added/removed/locked?

I'd have expected, if a document is removed from or added to the shared repository, then a repository that is shared between multiple appliction servers would reflect the same on each application server node? the only possible difference would be if some state specific data is persisted on the application servers (which enterprise/hazelcast then syncs).

I should have also said, that our architechure inlcudes an independant Solr server.

Thanks,
AN

Any thoughts mrogers?

Regards,
AN

mrogers
Star Contributor
Star Contributor
Yes it is the cached and ephemeral data that is the problem.     The application caches are important for performance. And some ephemeral data (like the application event queue) and lock status is not persisted at all.