Definitely not at a high level. RELAY will only have proper support when we implement the cross data centre feature in Infinispan for 5.2.x
Currently, you might find that info digging/extracting/hacking into the underlying JGroups code.
Galder: thanks for the response.
What are the changes related to RELAY (the cross data center feature) in 5.2 Alapha? I'm interesting to test it work.
What is release date for the 5.2 beta?
Dex, this is stuff Mircea is working on as part of https://issues.jboss.org/browse/ISPN-1824, design doc in:
He'll be able to tell you more. I've already fwd'd your previous questions to him but he's very busy right now.
Can't really tell when beta will be out but an alpha with a 1st version of this will be out (+-) beginning of August.
Btw, you can ping Mircea Markus on IRC if you're pressed with time....
Hi Dex and again sorry for the late answer.
The x-site replication from 5.1 was more experimental. Based on some initial feedback, we're working heavily right now on a newer version of cross-site replication based on a new implementation, RELAY2. As Galder pointed, you can find the design document with how this will work on the wiki. I imagine you have x-site scenario you want to handle - do yout think that this design coveres your scenarion?
Glader and Mircea, thanks for the responses.
Mircea: the user scenarios covered in the design doc (DOC-17546) covers our use cases.
I am glad to hear that you are actively work on this in 5.2. I'll be glad to help testing them when they are ready,
We are trying to use this feature for our geo-fail over deployment (either in active/passive (backup), or active/active, which is your use case #3).
Couple questions which I posted on other threads earlier, which I do not see mention in your DOC:
1) do you support Active/Hot standby with only one directional network accsss between 2 clusters?. Here is what I mean of one directional network access:
nodes in active cluster (local) can have access (TCP connection) to nodes in backup cluster, but not other way around. This is like a situation where local cluster inside a company's firewall, and standby cluster in Amazon EC2. Ususally, the firewall rules of a company allow nodes inside the compay to access nodes in Amazon EC2, but the nodes in EC2 are not able to access nodes inside company's fiew wall.
2) do you support to enable data replication across clusters for only certain name caches and not for others with a cache manager?
For example, I have 3 named caches cofigured, I only need to replication the data for one of them.
3) do you support the 2 cluster have diffent configurations such as cluster modes, numberOwners, etc?
For example, I have clustermode set to "REPLIACTION" in backup cluster and "DISTRIBUTION" in active cluster; have numberOwners=2 in active cluster and numberOwners=1 in backup cluster using "DIST" as cluster mode. Right now, hash is across all ndoes from 2 clusters. We should treat data distribution/replicatin with local cluster differently from across cluster.
Right now (5.1.5.Final), ISPN does not support #2 and #3 with RELAY.
Thanks for de fdeedback Dex!
- this cenario should be supported. It might get a bit trickier after the PASSIVE becomes ACTIVE and needs to backup its data to the site which is under a firewall - do you need to support that as well?
Re:2 & 3 the answer to both is yes.
In my view, there are 2 cases related to Active/Passive: a) Active /Passive and support fail back. In this case, yes, we have to require bi-directional (symetric) network access to support fail-over and fail-back.
b) Active/Passive one-direcctional network access. In this case, we do not support fail back (within ISPN). The passive site serves mainly for backup. As you stated that it will require out-bound mechanism to resync data back to Active site if a user wants to support fail-back. I would think this feature is more or less a nice to have.
What is the time line for those features (2 and 3 in my earlier list)? I'll love to try them out.