Seed node in a cluster

[netflix]

Seed node in a cluster

[Vijay2win]

Seed node is one of the node in the ASG... If you run it in multiple zones then it will one from each Zone.

The REST call will change, but cassandra needs the list only during the boot time after that it stores every nodes Address internally.

[unixboy]

what happens if the seen node dies in a cluster?

[Vijay2win]

ASG is suppose to spin another instance and Priam will recognize that the old node is not going to come up and will bootstrap the node with the old token... Cassandra will in-turn do its magic.

[unixboy]

if it is a 0 token on the node then will a new one with a 0 token come on line?

[Vijay2win]

Yep. make sure you have priam and cassandra baked into a AMI and you use SDB for metadata.

[unixboy]

so if a new sead token node somes on line then it will take over for the failed one?
does the other nodes in the ring need to be restarted to see the need seed node?

[unixboy]

ok I have had a bit more luck.

I just cant seem to get the cluster up and running I can see there is a seed node on one of the nodes but not any other one.

do I have to modify the cassandra.yaml file in the tomcat direcotry for the seed node ?

[Vijay2win]

https://github.com/Netflix/Priam/wiki/Setup this doesnt help? You dont need to change anything in Cassandra.yaml file Priam will overwrite it for you.... All you need is 2 Zones if possible.

[unixboy]

what is the cap for the number of servers in a zone? can you explane how the security groups are setup for the clusters?

[Vijay2win]

There isn't a cap on the number of servers....

We recommend you create Separate ASG's for each zone, it has more to do with the way Autoscalers work (if one node goes down in a zone and no other node is available in the ASG then it will spin a node in the other zone and will randomly kill a node in the zone which it created a node).
You just need to setup one security group for cassandra and let cassandra within a region use it.

If you have multi region cluster then you should be using Ec2MultiRegionSnitch, enable encription (in cassandra) and set the property so Priam can open the ports for communication.

[sagarl]

Looks like this issue is resolved, hence closing it.