Alfresco 3.2 Community Clustering Issues

[hailpam]

Hi all,

I'm trying to deploy in a production environment the Alfresco Cluster Configuration.

My ideas is the following:

1. the use of two CentOs linux servers, named A and B;
2. the use of NFS export to share SAN storage mount point (no replicated content store);
3. the use of MySql Cluster;
4. the use of a sticky load balancer in front of the two servers A and B;

in ordert to deploy an High Availability solution.

The problems encountered are those more common! Particularly, after the setting of

*. ehcache-custom.xml simple renaming the .sample file (no changes were made);
*. alfresco-global.properties to set the index tracker to sleep 2 seconds after every polling for lucene index refresh;
*. properties concerning the log of index tracker and ehcache;

I can see the heathbeats of ehcache and the PUT and REMOVE messages that server A and B exchange one with another (respectively, and according to the operations maked), but, and this is the strange thing, I don't see the logging messages concerning the scheduled Job of index tracker.

This last thing gives me to think! In fact, when I make the tests in order to understand if the cluster is correctly set, the only one test that I'm not able to overcome is just that concerning the lucene search: when I insert a file on server A, after that this operation succeed, I can see on B the same file, ok that's fine! Also, when I modify the attributes of file inserted on A, directly on A, after that this last operation suceed, I can see the modified fields on server B, and also this is fine! During the operations, on the log screen I can see the ehcache messages exchanged between the two servers.

The problems occurs when I try to do a lucene search on server B of the file that first I've created on server A. The search returns with 0 results, also after many hours from the operation of creation. Another strange thing that I can observe is the following: when I try to modify the attributes of the created file on server A from server B, I obtain a very long suspension of service of the logged session (web-browser is in 'waiting for' state… ), a so long suspension that I must abort the session, or with logout or with web-browser tab closure.

Can anyone help me to resolve this problem? I'm fighting with this problem from some days, and I'm tired of sever restarts every time I modify a simple parameter to try one alternative configuration in the set of the indefinite combinations of possibile configurations! 

If it needed I can post the configuration files (alfresco-gobal.properties and ehcache-custom.xml)…

Thanks in advance. Best Regards,

- Paolo

[derek]

Hi,
http://wiki.alfresco.com/wiki/Cluster_Configuration_V2.1.3_and_Later#High_availability_components

NB: As of V3.1, the cluster components will only initialize if the following property is defined:

alfresco.cluster.name=<CLUSTERNAME>

[hailpam]

Hi,
http://wiki.alfresco.com/wiki/Cluster_Configuration_V2.1.3_and_Later#High_availability_components

NB: As of V3.1, the cluster components will only initialize if the following property is defined:

alfresco.cluster.name=<CLUSTERNAME>

Hi derek,

thanks a lot for your reply: by the first tests I can observe that my problem has been resolved using your tip!

Best Regards,

- Paolo

P.S.: I'll deploy the cluster in the production environment after a set of test; I'll keep you informed!

[liferay01]

Hi all,

I'm trying to deploy in a production environment the Alfresco Cluster Configuration.

My ideas is the following:

1. the use of two CentOs linux servers, named A and B;
2. the use of NFS export to share SAN storage mount point (no replicated content store);
3. the use of MySql Cluster;
4. the use of a sticky load balancer in front of the two servers A and B;

in ordert to deploy an High Availability solution.
- Paolo

Hi derek,

thanks a lot for your reply: by the first tests I can observe that my problem has been resolved using your tip!

Best Regards,

- Paolo

P.S.: I'll deploy the cluster in the production environment after a set of test; I'll keep you informed!

Hi Paolo,

We are on a similar situation. We are trying to buid a high availability solution based on :

1. the use of two Ubuntu linux servers;
2. a replicated content store;
3. the use of MySql Cluster;
4. the use of a sticky load balancer in front of the two servers;

We are facing some problems with the MySQL Cluster. What we are doing is

1. Extract the database schema made for MySQL InnoDB engine
2. Modify the extracted file to adapt it to the NDBCluster engine (replace InnoDB with NDBCluster)
3. Import the database schema

Unfortunately, this leads to the following errors :
ERROR 1005 (HY000) at line 1126: Can't create table 'alfresco.alf_access_control_entry' (errno: 906)
by the way, "ERRNO 906" is "DMEC SE Unsupported attribute type in index"

Reading your post, It looks like you succeded using the MySQL Cluster with Alfresco. Is that rigth ?
If so, how did you manage to put the alfresco database schema in the MySQL Cluster without errors ?

Regards

[hailpam]

Hi liferay01,

sorry for a such delayed reply to your question.

I encountered your same problem with MySql cluster DB engine (NDBcluster), but I've not been able to solve it following the solution you have proposed.
There is a persistence control at Alfresco's ORM layer that is blocking for DB schema deploy that is different from InnoDB. The error errno:906 concerning the index creation, and can be caused by the above mentioned blocking control: "Can't create table… ", Alfresco tries to create InnoDB tables that are serialized on local datafile regardless the already existing NDBcluster tables that instead are 'in-memory entities'.

Regars,

- Paolo