The Seal installation path needs to be accessible to all the nodes in your Hadoop cluster. In addition, Seal’s runtime dependencies also need to be accessible to all cluster nodes.
Deployment strategies for Seal on your cluster will vary depending on your particular set-up.
For large clusters, it may be necessary to distribute the software to the various nodes rather than relying on a shared volume. In our tests, a volume shared via NFS was not able to handle the demands of a cluster with 128 nodes, resulting in tasks that never went past the “initializing” state and eventually timed out.
If you have a cluster this large, you should probably ask your system administrators for advice on how to install Seal.