Quick Start Improve this page

Quick Start: Build a star detection pipeline

This quick start guide will demonstrate how to:

  • create, configure, and interact with OpenStack cloud and Virtual Machines (VMs)
  • store files with the CANFAR VOSpace storage
  • launch batch processing jobs from the CANFAR batch host, using VMs created in the previous step

Before starting this example, you will need to register to CANFAR. The CANFAR team will take care of your registration to Compute Canada infrastructure.

Create your interactive Virtual Machine

To manage the VMs with OpenStack, we suggest using the dashboard at Compute Canada. Log into the dashboard. Provide your CANFAR username, adding a -canfar suffix, e.g, janesmith-canfar, and your usual CANFAR password. We will refer the CANFAR username (excluding the -canfar suffix which is only used for logging into the dashboard) as [username] throughout this document.

Each resource allocation corresponds to an OpenStack tenant or project (these two names are used interchangeably). A user may be a member of multiple tenants, and a tenant usually has multiple users. A pull-down menu near the top-left allows you to select between the different tenants that you are a member of.

Allow ssh access to your VM

  • Click on Access & Security (left column of page), and then the Security Groups tab.
  • Click on the Manage Rules button next to the default group. If you see a rule with Ingress direction, 22(SSH) Port Range and 0.0.0.0/0 (CIDR) , then that means someone in your tenant already opened the ssh port for you. If you don’t see it, add a new rule following step.
  • Click on the + Add Rule button near the top-right. Select SSH at the bottom of the Rule pull-down menu, and then click on Add at the bottom-right. This operation is only required once for the initial setup of the tenant.

Import an ssh public key

Access to VMs is facilitated by SSH key pairs rather than less secure user name / password. A private key resides on your own computer, and the public key is copied to all machines that you wish to connect to.

  • If you have not yet created a key pair on your system, run ssh-keygen on this local machine to generate one or follow this documentation for example.
  • Click on Access & Security, switch to the Key Pairs tab and click on the Import Key Pair button at the top-right.
  • Choose a meaningful name for the key, and then copy and paste the contents of ~/.ssh/id_rsa.pub from the machine you plan to ssh from into the Public Key window.

Allocate a public IP address

You will need to connect to your VM via a public IP.

  • Click on the Floating IPs tab. If there are no IPs listed, click on the Allocate IP to Project button at the top-right.

Each tenant/project will typically have one public IP. If you have already allocated all of your IPs, this button will read “Quota Exceeded”.

Launch a VM

We will now launch a VM with Ubuntu 16.04.

  • Switch to the Images window (left-hand column), and then click on the Public button at top right (it might be already selected.
  • Select ubuntu-server-16.04-amd64 and click on the Launch Instance button on the right.
  • In the Details tab choose a meaningful Instance Name. Flavor is the hardware profile for the VM. c2-7.5gb-80 provides the minimal requirements of 2 cores, 7.5GB or RAM for most VMs. Note that it provides an 80 GB ephemeral disk that will be used as scratch space for batch processing. Availability Zone should be left empty, and Instance Count 1.
  • In the Access & Security tab ensure that your public key is selected, and the default security group (with ssh rule added) is selected.
  • Finally, click the Launch button.

Connect to the VM

After launching the VM you are returned to the Instances window. You can check the VM status once booted by clicking on its name (the Console tab of this window provides a basic console in which you can monitor the boot process). Before being able to ssh to your instance, you will need to attach the public IP address to it.

  • Select Associate Floating IP from the More pull-down menu.
  • Select the address that was allocated and the new VM instance in the Port to be associated menu, and click on Associate.

Your ssh public key will have been injected into a generic account with a name like centos, or ubuntu, depending on the Linux distribution. To discover the name of this account, first attempt to connect as root:

ssh root@[floating ip]
Please login as the user "ubuntu" rather than the user "root".


ssh ubuntu@[floating ip]

Create a user on the VM

You might need to create a different user than the default one, and for batch processing to work, it is presently necessary for you to create a user on the VM with your CANFAR username. You can use a wrapper script for this:

curl https://raw.githubusercontent.com/canfar/openstack-sandbox/master/scripts/canfar_create_user.bash -o canfar_create_user.bash
sudo bash canfar_create_user.bash [username]

Now, exit the VM, and re-connect with your CANFAR username instead of the standard ubuntu username:

exit
ssh [username]@[floating_ip]

Install software on the VM

The VM operating system has only a minimal set of packages. For this example, we will use:

  • the SExtractor package to create catalogues of stars and galaxies.
  • We also need to read FITS images. Most FITS images from CADC come Rice-compressed with an fz extension. SExtractor only reads uncompressed images, so we also need the funpack utility to uncompress the incoming data. The funpack executable is included in the package libcfitsio-bin.

Let’s install them both system wide after a fresh update of the Ubuntu repositories:

sudo apt update -y
sudo apt install -y sextractor libcfitsio-bin

Test the Software on the VM

We are now ready to do a simple test. Let’s download a FITS image to our scratch space. When we instantiated the VM we chose a flavour with an ephemeral disk. First, execute the following script to mount this device at /ephemeral and create a work directory to mimic the batch processing environment (note that this will be done automatically for batch jobs):

curl https://raw.githubusercontent.com/canfar/openstack-sandbox/master/scripts/canfar_mount_ephemeral.bash -o canfar_mount_ephemeral.bash
sudo bash canfar_mount_ephemeral.bash
sudo mkdir /ephemeral/work
sudo chown [username]:[username] /ephemeral/work

Next, enter the directory, copy an astronomical image there, and run SExtractor on it:

cd /ephemeral/work
cp /usr/share/sextractor/default* .
rm default.param
echo 'NUMBER
MAG_AUTO
X_IMAGE
Y_IMAGE'
> default.param
curl -L http://www.canfar.phys.uvic.ca/data/pub/CFHT/1056213p | funpack -O 1056213p.fits -
sextractor 1056213p.fits -CATALOG_NAME 1056213p.cat

The image 1056213p.fits is a Multi-Extension FITS file with 36 extensions, each containing data from one CCD from the CFHT Megacam camera.

Store results on the CANFAR VOSpace

All data stored on the VM and ephemeral disk since the last time it was saved are normally wiped out when the VM shuts down). We will use the CANFAR VOSpace to store the result. We want to store the output catalogue 1056213p.cat at a persistent, externally-accessible location. We will use the CANFAR VOSpace for this purpose. To store anything on the CANFAR VOSpace from a command line, you will need the CANFAR VOSpace client.

sudo apt install -y python-pip
sudo pip install vos

For an automated procedure to access VOSpace on your behalf, your proxy authorization must be present on the VM. You can accomplish this using a .netrc file that contains your CANFAR user name and password, and the command getCert can obtain an X509 Proxy Certificate using that username/password combination without any further user interaction. The commands below will create the file and install the VOSpace utilities.

echo "machine www.canfar.phys.uvic.ca login [username] password [password]" > ~/.netrc
chmod 600 ~/.netrc
getCert

Let’s check that the VOSpace client works by copying the results to your VOSpace:

vcp 1056213p.cat vos:[username]

Verify that the file is properly uploaded by pointing your browser to the VOSpace browser interface.

Automate all the above and run it in batch

Now we want to automate the whole procedure above in a single script, in preparation for batch processing. Paste the following commands into one BASH script named ~/do_catalog.bash on your VM:

#!/bin/bash
source /home/[username]/.bashrc
curl -L http://www.canfar.phys.uvic.ca/data/pub/CFHT/${1} | funpack -O ${1}.fits -
cp /usr/share/sextractor/default* .
echo 'NUMBER
MAG_AUTO
X_IMAGE
Y_IMAGE'
> default.param
sextractor ${1}.fits -CATALOG_NAME ${1}.cat
getCert
vcp ${1}.cat vos:[username]

Remember to substitute [username] with your CANFAR user account.

This script runs all the commands, one after the other, and takes only one parameter represented by by the shell variable ${1}, the file ID of the CFHT exposure. Save your script and set it as executable:

chmod +x ~/do_catalog.bash

Now let’s test the newly created script with a different file ID. The do_catalog.bash script will run on the local directory where it is launched from. Let’s emulate a batch job and launch it from the ephemeral disk:

cd /ephemeral/work
~/do_catalog.bash 1056214p

Just as we did in the previous manual test, verify the output, and check with the VOSpace web interface that the catalogue has been uploaded.

Finally, make a copy of the script on your local machine so that it will be available for submitting batch jobs once the VM is shut down, e.g.,

scp [username]@[floating_ip]:do_catalog.bash .

Install HTCondor for Batch

Batch jobs are scheduled using a software package called HTCondor. HTCondor will dynamically launch jobs on the VMs (workers), connecting to the batch processing head node (the central manager). In order to install HTCondor (which provides a minimal HTCondor daemon to execute jobs) run this script on your VM instance:

curl https://raw.githubusercontent.com/canfar/openstack-sandbox/master/vm_config/canfar_batch_setup.bash -o canfar_batch_setup.bash
sudo bash canfar_batch_setup.bash

Snapshot (save) the VM Instance

Now we want to save our work. Return to your browser on the OpenStack dashboard.

  • Save the state of your VM by navigating to the Instances window of the dashboard, and click on the Create Snapshot button to the right of your VM instance’s name. After selecting a name for the snapshot of the VM instance, e.g., my_vm_image, click the Create Snapshot button. It will eventually be saved and listed in the VM Images window, and will be available next time you launch an instance of that VM image.

Shutdown the VM Instance

  • In the Instances window, select Terminate Instance in the More pull-down menu, and confirm.

We are now are ready to launch batch processing jobs creating catalogues of many CFHT Megacam images and uploading the catalogues to VOSpace.

Write your batch processing job submission

Assuming you have the do_catalog.bash script written above on your local desktop, copy it to the CANFAR batch host, and then log into the batch host:

scp docatalog.bash [username]@batch.canfar.net:
ssh [username]@batch.canfar.net

Let’s write a submission file that will transfer the do_catalog.bash script to the execution host. The execution host will be an instance of your snapshot VM image with 4 cores, and for each given CADC CFHT file id, will run a job on one of the core. The job will consist of running your script for 4 CFHT images with the file IDs 1056215p, 1056216p, 1056217p, and 1056218p. For this tutorial you will modify the configuration file listed below. Fire up your favorite editor and paste the following text into a submission file:

Universe   = vanilla
should_transfer_files = YES
when_to_transfer_output = ON_EXIT_OR_EVICT
environment = "HOME=/home/[username]"
RunAsOwner = True

transfer_output_files = /dev/null

Executable = do_catalog.bash

Arguments = 1056215p
Log = 1056215p.log
Output = 1056215p.out
Error = 1056215p.err
Queue

Arguments = 1056216p
Log = 1056216p.log
Output = 1056216p.out
Error = 1056216p.err
Queue

Arguments = 1056217p
Log = 1056217p.log
Output = 1056217p.out
Error = 1056217p.err
Queue

Arguments = 1056218p
Log = 1056218p.log
Output = 1056218p.out
Error = 1056218p.err
Queue

Again, be sure to substitue the correct value for [username]. It is important to set this HOME environment variable so that the running job will be able to locate the .netrc file with VOSpace credentials.

Save the submission file as myjobs.sub.

Submit the batch jobs

Source the OpenStack RC tenant file, and enter your CANFAR password. This sets environment variables used by OpenStack (only required once per login session):

. canfar-[tenant]-openrc.sh
Please enter your OpenStack Password:

You can then submit your jobs to the condor job pool:

canfar_submit myjobs.sub my_vm_image c4-15gb-83

my_vm_image is the name of the snapshot you used during the VM configuration above, and c4-15gb-83 is the flavor for the VM(s) that will execute the jobs. If you wish to use a different flavor, they are visible through the dashboard when launching an instance, or using the nova command-line client.

After submitting, wait a couple of minutes. Check where your jobs stand in the queue:

condor_q [username]

Check the status of the VMs and jobs running on the cloud:

condor_status -submitter

Once you have no more jobs in the queue, check the logs and output files myjobs.* on the batch host, and check on your VOSpace browser. All 4 of the generated catalogues should have been uploaded.

You are done!