Configuration Suggestions for Multiple Datacenters & EC2

[OzgurCD]

Hello,

What configuration steps would you suggest if I want to create a
hazelcast cluster on multiple datacenters & EC2 hybrid structure.
Using EC2 auto-discovery options is not applicable in our case.

My aim is to have something like; two hosts in each datacenter (and
EC2) as a bridge (to use their IP's in join statements in config) and
use a datacenter specific configuration for hosts on each datacenter.

I hope my question make sense (:

Thank you,
Ozgur

[Andrei Pozolotin]

I use aws simpledb hz node registrations in this case

[Talip Ozturk]

Merhaba Ozgur,

Let me first check if I got your purpose right:
Let's say you have 10 nodes in datacenter1 and another 10 nodes in
datacenter2 and 5 nodes in ec2. You want to create a cluster of 25
nodes; all sharing the same distributed map, for example.
If so, can you tell me why you want it like that? As you might know
Hazelcast is peer2peer clustering and all nodes are connected to each
other. Quality of network is so important. I would be worried about
the likelihood of disconnection between datacenters (with ec2).

Tell us a little more about your environment and purpose so that we
can possibly give you a better answer.

http://twitter.com/oztalip

> --
> You received this message because you are subscribed to the Google Groups "Hazelcast" group.
> To post to this group, send email to haze...@googlegroups.com.
> To unsubscribe from this group, send email to hazelcast+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/hazelcast?hl=en.
>
>

[OzgurCD]

Selam Talip,

Yes, that's correct. This is exactly what I need to have. I want to
use Hazelcasts distributed executor & queue features and I believe,
in my case the concerns you mentioned are not big deals, can be
ignored.

Thank you

[Ed Anuff]

I'm curious about this as well. When I looked at the docs, it seemed
that this sort of configuration was possible. Is that not the case?

[Fuad Malikov]

Ed,

By "this sort of configuration' what do you mean?

If you mean Hazelcast across multiple datacenters? That is OK, but Hazelcast requires all nodes to have connection with each other? And operations are sensitive to the latency. You may not get the best out of Hazelcast. Instead you can think of having seperate clusters in each data center and syncing them. The current code at trunk has the WAN replication. But you'll need to wait a bit for the details to come out.

Fuad

--
Fuad Malikov
+90.538.378.9777
http://twitter.com/fuadm

[Ed Anuff]

I meant having a cluster that spans datacenters. Would it make more
sense to have applications in datacenter B connect to a cluster that
lives in datacenter A as native clients rather than having a cluster
span datacenters? I'm specifically thinking about for things like
locking. If the cluster was unreachable by an app in datacenter B,
the app know that it couldn't perform certain operations safely and
act accordingly.

Ed