cloudtools is a small collection of command line tools intended to make using Hail on clusters running in Google Cloud's Dataproc service simpler.
These tools are written in Python and mostly function as wrappers around the gcloud suite of command line tools included in the Google Cloud SDK.
Prerequisites:
- Mac OS X
- Python 2 or 3
- Google Cloud SDK
- (Optional) Google Chrome installed in the (default) location
/Applications/Google Chrome.app/Contents/MacOS/Google Chrome
cloudtools can be installed from the Python package index using the pip installer: pip install cloudtools
To update to the latest version: pip install cloudtools --upgrade
All functionality in cloudtools is accessed through the cluster module.
There are 6 commands within the cluster module:
cluster start <name> [args]cluster submit <name> [args]cluster connect <name> [args]cluster diagnose <name> [args]cluster stop <name>cluster list
where <name> is the required, user-supplied name of the Dataproc cluster.
REMINDER: Don't forget to shut down your cluster when you're done! You can do this using cluster stop <name>, through the Google Cloud Console, or using the Google Cloud SDK directly with gcloud dataproc clusters delete name.
One way to use the Dataproc service is to write complete Python scripts that use Hail, and then submit those scripts to the Dataproc cluster. An example of using cloudtools to interact with Dataproc in this way would be:
$ cluster start testcluster -p 6
...wait for cluster to start...
$ cluster submit testcluster myhailscript.py
...Hail job output...
Job [...] finished successfully.
where myhailscript.py lives on your computer in your current working directory and looks something like:
import hail as hl
hl.init()
...
This snippet starts a cluster named "testcluster" with the 1 master machine, 2 worker machines (the minimum/default), and 6 additional preemptible worker machines. Then, after the cluster is started (this can take a few minutes), a Hail script is submitted to the cluster "testcluster".
You can also pass arguments to the Hail script using the --args argument:
$ cluster submit testcluster myhailscript.py --args "arg1 arg2"
where myhailscript.py is
import sys
print('First argument: ', sys.argv[1])
print('Second argument: ', sys.argv[2])
would print
First argument: arg1
Second argument: arg2
Another way to use the Dataproc service is through a Jupyter notebook running on the cluster's master machine. By default, cluster name start sets up and starts a Jupyter server process - complete with a Hail kernel - on the master machine of the cluster.
To use Hail in a Jupyter notebook, you'll need to have Google Chrome installed on your computer as described in the installation section above. Then, use
cluster connect testcluster notebook
to open a connection to the cluster "testcluster" through Chrome.
A new browser will open with the address localhost:8123 -- this is port 8123 on the cluster's master machine, which is where the Jupyter notebook server is running. You should see the Google Storage home directory of the project your cluster was launched in, with all of the project's buckets listed.
Select the bucket you'd like to work in, and you should see all of the files and directories in that bucket. You can either resume working on an existing .ipynb file in the bucket, or create a new Hail notebook by selecting Hail from the New notebook drop-down in the upper-right corner.
From the notebook, you can use Hail the same way that you would in a complete job script:
import hail as hl
hl.init()
...
To read or write files stored in a Google bucket outside of Hail-specific commands, use Hail's hadoop_read() and hadoop_write() helper functions. For example, to read in a TSV file from Google storage to a pandas dataframe:
import hail as hl
import pandas as pd
hl.init()
with hl.hadoop_open('gs://mybucket/mydata.tsv', 'r') as f:
df = pd.read_table(f)
When you save your notebooks using either File -> Save and Checkpoint or command + s, they'll be saved automatically to the bucket you're working in.
While your job is running, you can monitor its progress through the Spark Web UI running on the cluster's master machine at port 4040. To connect to the SparkUI from your local machine, use
cluster connect testcluster ui
If you've attempted to start multiple Hail/Spark contexts, you may find that the web UI for a particular job is accessible through ports 4041 or 4042 instead. To connect to these ports, use
cluster connect testcluster ui1
to connect to 4041, or
cluster connect testcluster ui2
to connect to 4042.
To view details on a job that has completed, you can access the Spark history server running on port 18080 with
cluster connect testcluster spark-history
$ cluster -h
usage: cluster [-h] {start,submit,connect,diagnose,stop} ...
Deploy and monitor Google Dataproc clusters to use with Hail.
positional arguments:
{start,submit,connect,diagnose,stop}
start Start a Dataproc cluster configured for Hail.
submit Submit a Python script to a running Dataproc cluster.
connect Connect to a running Dataproc cluster.
diagnose Diagnose problems in a Dataproc cluster.
stop Shut down a Dataproc cluster.
optional arguments:
-h, --help show this help message and exit
$ cluster start -h
usage: cluster start [-h] [--hash HASH] [--spark {2.0.2,2.2.0}]
[--version {0.1,0.2}]
[--master-machine-type MASTER_MACHINE_TYPE]
[--master-memory-fraction MASTER_MEMORY_FRACTION]
[--master-boot-disk-size MASTER_BOOT_DISK_SIZE]
[--num-master-local-ssds NUM_MASTER_LOCAL_SSDS]
[--num-preemptible-workers NUM_PREEMPTIBLE_WORKERS]
[--num-worker-local-ssds NUM_WORKER_LOCAL_SSDS]
[--num-workers NUM_WORKERS]
[--preemptible-worker-boot-disk-size PREEMPTIBLE_WORKER_BOOT_DISK_SIZE]
[--worker-boot-disk-size WORKER_BOOT_DISK_SIZE]
[--worker-machine-type WORKER_MACHINE_TYPE] [--zone ZONE]
[--properties PROPERTIES] [--metadata METADATA]
[--packages PACKAGES] [--jar JAR] [--zip ZIP]
[--init INIT] [--init_timeout INIT_TIMEOUT] [--vep] [--dry-run]
name
Start a Dataproc cluster configured for Hail.
positional arguments:
name Cluster name.
optional arguments:
-h, --help show this help message and exit
--hash HASH Hail build to use for notebook initialization
(default: latest).
--spark {2.0.2,2.2.0}
Spark version used to build Hail (default: 2.2.0)
--version {0.1,0.2}
Hail version to use (default: 0.2).
--master-machine-type MASTER_MACHINE_TYPE, --master MASTER_MACHINE_TYPE, -m MASTER_MACHINE_TYPE
Master machine type (default: n1-highmem-8).
--master-memory-fraction MASTER_MEMORY_FRACTION
Fraction of master memory allocated to the JVM. Use a
smaller value to reserve more memory for Python.
(default: 0.8)
--master-boot-disk-size MASTER_BOOT_DISK_SIZE
Disk size of master machine, in GB (default: 100).
--num-master-local-ssds NUM_MASTER_LOCAL_SSDS
Number of local SSDs to attach to the master machine
(default: 0).
--num-preemptible-workers NUM_PREEMPTIBLE_WORKERS, --n-pre-workers NUM_PREEMPTIBLE_WORKERS, -p NUM_PREEMPTIBLE_WORKERS
Number of preemptible worker machines (default: 0).
--num-worker-local-ssds NUM_WORKER_LOCAL_SSDS
Number of local SSDs to attach to each worker machine
(default: 0).
--num-workers NUM_WORKERS, --n-workers NUM_WORKERS, -w NUM_WORKERS
Number of worker machines (default: 2).
--preemptible-worker-boot-disk-size PREEMPTIBLE_WORKER_BOOT_DISK_SIZE
Disk size of preemptible machines, in GB (default:
40).
--worker-boot-disk-size WORKER_BOOT_DISK_SIZE
Disk size of worker machines, in GB (default: 40).
--worker-machine-type WORKER_MACHINE_TYPE, --worker WORKER_MACHINE_TYPE
Worker machine type (default: n1-standard-8, or
n1-highmem-8 with --vep).
--zone ZONE Compute zone for the cluster (default: us-central1-b).
--properties PROPERTIES
Additional configuration properties for the cluster
--metadata METADATA Comma-separated list of metadata to add:
KEY1=VALUE1,KEY2=VALUE2...
--packages PACKAGES, --pkgs PACKAGES
Comma-separated list of Python packages to be
installed on the master node.
--jar JAR Hail jar to use for Jupyter notebook.
--zip ZIP Hail zip to use for Jupyter notebook.
--init INIT Comma-separated list of init scripts to run.
--init_timeout INIT_TIMEOUT
Flag to specify a timeout period for the
initialization action
--vep Configure the cluster to run VEP.
--dry-run Print gcloud dataproc command, but don't run it.```
$ cluster submit -h
usage: cluster submit [-h] [--properties PROPERTIES]
[--args ARGS]
name script
Submit a Python script to a running Dataproc cluster.
positional arguments:
name Cluster name.
script
optional arguments:
-h, --help show this help message and exit
--properties PROPERTIES, -p PROPERTIES
Extra Spark properties to set.
--args ARGS Quoted string of arguments to pass to the Hail script
being submitted.
$ cluster connect -h
usage: cluster connect [-h] [--port PORT] [--zone ZONE]
name
{notebook,nb,spark-ui,ui,spark-ui1,ui1,spark-ui2,ui2,spark-history,hist}
Connect to a running Dataproc cluster.
positional arguments:
name Cluster name.
{notebook,nb,spark-ui,ui,spark-ui1,ui1,spark-ui2,ui2,spark-history,hist}
Web service to launch.
optional arguments:
-h, --help show this help message and exit
--port PORT, -p PORT Local port to use for SSH tunnel to master node
(default: 10000).
--zone ZONE, -z ZONE Compute zone for Dataproc cluster (default: us-
central1-b).
$ cluster diagnose -h
usage: cluster diagnose [-h] --dest DEST [--hail-log HAIL_LOG] [--overwrite]
[--no-diagnose] [--compress]
[--workers [WORKERS [WORKERS ...]]] [--take TAKE]
name
Diagnose problems in a Dataproc cluster.
positional arguments:
name Cluster name.
optional arguments:
-h, --help show this help message and exit
--dest DEST, -d DEST Directory for diagnose output -- must be local.
--hail-log HAIL_LOG, -l HAIL_LOG
Path for hail.log file.
--overwrite Delete dest directory before adding new files.
--no-diagnose Do not run gcloud dataproc clusters diagnose.
--compress, -z GZIP all files.
--workers [WORKERS [WORKERS ...]]
Specific workers to get log files from.
--take TAKE Only download logs from the first N workers.
$ cluster stop -h
usage: cluster stop [-h] name
Shut down a Dataproc cluster.
positional arguments:
name Cluster name.
optional arguments:
-h, --help show this help message and exit
TWINE_USERNAME=username TWINE_PASSWORD=password make deploy
pypi-username and pypi-password are files containing credentials sufficiently privileged to publish cloudtools
kubectl create secret generic \
ci-deploy-0-1--nealelab-cloudtools \
--from-file=secrets/pypi-username \
--from-file=secrets/pypi-password