Synchronous distributed and backup

[lucdew]

Hi,

I am evaluation infinispan 6.0 and I have questions regarding the synchronous distributed cluster mode.

I did some tests with 4 nodes synchronizing with JGroups in tcp. I put 800 Mbytes of data in the cluster (each VM started with > 1Gbytes), then stopped adding any data.

If a node was lost, I could still access any entry of the cache from any of the remaining nodes.

However as soon as I lost another node with 2 remaining, some entries were lost.

What is the backup strategy in the distributed mode ? Is there any backup at all (if you say no I think i have to review how my test was implemented ;-) )

Can the number of backup nodes when an entry is added be configured ? ( I am not referring to near cache).

Regards,

Luc

[wdfink]

With distributed mode you can set the number of 'owners' which mean how many nodes should store an entity.

If a cluster node fail there is a re-hashing algorithm and the remaining members copy the entries (where the failed node was the owner) to an existing node.

You might attach your configuration and more details.