For the ones that need access by public IP we have not found a way to automate it. Would be curious to know if anyone else has been able to that.
In the case of access by private IP we just specify security group as the source.
From: Alain RODRIGUEZ [mailto:email@example.com]
Sent: Wednesday, June 05, 2013 5:45 PM
Subject: Re: Looking for a fully working AWS multi DC configuration.
Do you open all these nodes one by one on every Security Group in each region every time you add a node or did you manage to automate it somehow ?
2013/6/5 Dan Kogan <firstname.lastname@example.org>
We are using a very similar configuration. From our experience, Cassandra nodes in the same DC need access over both public and private IP on the storage port (7000/7001). Nodes from other DC will need access over public IP on the storage port.
All Cassandra nodes also need access over the public IP on the Thrift port (9160).
We use to work on a single DC (EC2Snitch / SimpleStrategy). For latency reason we had top open a new DC in the US (us-east). We run C* 1.2.2. We don't use VPC.
Now we use:
- 2 DC (eu-west, us-east)
- EC2MultiRegionSnitch / NTS
- public IPs as broadcast_address and seeds
- private IPs as listen_address
Yet we are experimenting some troubles (node can't reach itself, Could not start register mbean in JMX...), mainly because of the use of public IPs and the AWS inter-region communication.
If someone has successfully setup this kind of cluster, I would like to know, if our configuration is correct and if I am missing something.
I also would like to know what ports I have to open and either where I have to open them from.
Any insight would be greatly appreciated.