4
- Apache Cassandra / Spark on Kubernetes
5
While it is more common to run Kubernetes on top of Mesos cluster (and Mesos
6
support is directly integrated in Kubernetes 1.2 which will work as another
7
Messos scheduler; it is not clear, however, if this enabled in OpenShift's
8
version of Kubernetes), a few people also run Apache stack in the containers.
9
* The solution is to run ZooKeeper and other services as Kubernetes DaemonSets
10
and use selectors if they should be deployed on subset of nodes.
11
* For performance reason, instead of shared Gluster storage, the hostPath
12
volumes has to be used for data store. Cassandra has it's own replication
13
mechanism and using shared storage will just inflict significant performance
14
penalty. As we lock Cassandra to specific nodes, relying on host storage is
16
* Links (feature request & instructions):
17
https://github.com/kubernetes/kubernetes/issues/5017
18
http://blog.kubernetes.io/2016/03/using-Spark-and-Zeppelin-to-process-Big-Data-on-Kubernetes.html
21
There is several ways how to run MySQL. The simplest is to create a standard
22
MySQL container and schedule it run exactly on one host using Gluster for
23
data store. The faster option, as there is no penalty from shared storage,
24
to run 2 MySQL servers on specific nodes (using selectors) and the local
25
hostPath volumes. Then, either cache for ADEI is created independently on
26
both systems or Galera is used for replication.