kubernetes: Cassandra example on AWS doesn't create a Cassandra cluster

I’m running a simple cluster (k8s 1.1.1) on aws and trying to follow the instructions provided in the Cassandra examples section (https://github.com/kubernetes/kubernetes/tree/release-1.1/examples/cassandra).

When I try to scale the Cassandra nodes:

kubectl scale rc cassandra --replicas=2

the new pod is created as expected:

NAME              READY     STATUS    RESTARTS   AGE
cassandra         1/1       Running   0          21m
cassandra-yr4nq   1/1       Running   0          20m

but if I run:

kubectl exec -ti cassandra – nodetool status

I can see only one node (of course this is specular, so the same happens if I try querying the other pod, cassandra-yr4nq)

Datacenter: datacenter1
=======================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address     Load       Tokens  Owns (effective)  Host ID                               Rack
UN  10.244.0.8  51.49 KB   256     100.0%            2db6bea7-31df-436e-827c-87dca466d31d  rack1

Two more info:

I’ve also tried to check if the two pods can effectively see each other (logged in and tested with a ping) and they can
before going with AWS I tested this configuration on my local pc with vagrant and everything was fine (I was able to scale up and down the Cassandra nodes)

Any ideas?

About this issue

Original URL
State: closed
Created 9 years ago
Comments: 24 (14 by maintainers)

Most upvoted comments

t2.micro are WAY small 😦 I am bumping the demo mem up to 2G and 2CPU. I am not sure if the jvm will even start. I am pretty sure it will start with 1G of mem, but have not tried. Can you get any logs from kubectl.

Btw info on validating an install https://github.com/kubernetes/kubernetes/issues/20961#issuecomment-212147456 - but this is when they are running

I have noticed that nodetool will run SUPER slow with the current demo setup. You bump the mem and the cpu up, and it is much happier.

chrislovecnm on May 3, 2016