[SPARK-5189] Reorganize EC2 scripts so that nodes can be provisioned independent of Spark master

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Component/s: EC2
• Assignee: Unassigned
• Reporter: Nicholas Chammas
• Created: 10/Jan/15 19:42
• Updated: 27/Jan/16 13:25
• Resolved: 27/Jan/16 10:13

Description

As of 1.2.0, we launch Spark clusters on EC2 by setting up the master first, then setting up all the slaves together. This includes broadcasting files from the lonely master to potentially hundreds of slaves.

There are 2 main problems with this approach:

1. Broadcasting files from the master to all slaves using copy-dir (e.g. during ephemeral-hdfs init, or during Spark setup) takes a long time. This time increases as the number of slaves increases.
   I did some testing in us-east-1. This is, concretely, what the problem looks like:

   number of slaves (m3.large)	launch time (best of 6 tries)
   1	8m 44s
   10	13m 45s
   25	22m 50s
   50	37m 30s
   75	51m 30s
   99	1h 5m 30s

   Unfortunately, I couldn't report on 100 slaves or more due to SPARK-6246, but I think the point is clear enough.
   We can extrapolate from this data that every additional slave adds roughly 35 seconds to the launch time (so a cluster with 100 slaves would take 1h 6m 5s to launch).

2. It's more complicated to add slaves to an existing cluster (a la SPARK-2008), since slaves are only configured through the master during the setup of the master itself.

Logically, the operations we want to implement are:

• Provision a Spark node
• Join a node to a cluster (including an empty cluster) as either a master or a slave
• Remove a node from a cluster

We need our scripts to roughly be organized to match the above operations. The goals would be:

1. When launching a cluster, enable all cluster nodes to be provisioned in parallel, removing the master-to-slave file broadcast bottleneck.
2. Facilitate cluster modifications like adding or removing nodes.
3. Enable exploration of infrastructure tools like Terraform that might simplify spark-ec2 internals and perhaps even allow us to build one tool that launches Spark clusters on several different cloud platforms.

More concretely, the modifications we need to make are:

• Replace all occurrences of copy-dir or rsync-to-slaves with equivalent, slave-side operations.
• Repurpose setup-slave.sh as provision-spark-node.sh and make sure it fully creates a node that can be used as either a master or slave.
• Create a new script, join-to-cluster.sh, that takes a provisioned node, configures it as a master or slave, and joins it to a cluster.
• Move any remaining logic in setup.sh up to spark_ec2.py and delete that script.

Issue Links

• blocks: SPARK-2008 Enhance spark-ec2 to be able to add and remove slaves to an existing cluster
• Is contained by: SPARK-4325 Improve spark-ec2 cluster launch times