Skip to content

Frequently asked questions

What is the CANFAR Science Platform?

CANFAR provides interactive and unattended computing Sessions for astronomy, with access to project storage, configured VOSpace Services, and published Container Images. The available Servers, images, resources, and Storage Identifiers depend on the deployment and your account.

How do I get access?

You need a CADC account and access to a Science Platform Server. Request a CADC account if needed. Then either ask your project's administrator to add you to its group, or email support@canfar.net with your CADC username and a short description of your research. See Getting started.

Who can use CANFAR and what does it cost?

The CADC deployment serves Canadian astronomers and their collaborators. Access is provided for astronomical research, under the site's policies and within the storage and compute allocated to your project; it is not a general-purpose free cloud. A project that needs more than CANFAR can offer should apply for an Alliance resource allocation. An SRCNet Science Platform Server has its own eligibility rules and support contacts.

How do I log in and select a server?

For browser work, sign in to the Science Portal and launch a Session. You do not need to install the client.

For command-line or Python work, install the client and select your server:

canfar login cadc
canfar server ls
canfar server use SERVER_NAME
canfar config get active.server

canfar auth manages saved Authentication Records and canfar server manages the active Server. The removed canfar context command is not part of the current CLI.

Which Session Kind should I use?

  • notebook for interactive Python and exploratory work;
  • desktop for a graphical Linux environment;
  • carta for image and cube visualisation;
  • firefly for supported browser-based astronomy applications; and
  • headless for scripts, reductions, and parameter sweeps.

Use canfar image ls --kind KIND to see images available for a kind. See Sessions.

What does Pending mean?

Pending means the platform accepted the request but has not made the Session ready. Admission, requested resources, image pulling, or initialization can all occur while a Session is Pending. canfar create returns accepted IDs; it does not wait for Running.

In the browser, check the Session in Active Sessions and wait for its application link. If it remains pending, record its ID and contact support. With the client installed, collect more details:

canfar ps --all
canfar info SESSION_ID
canfar events SESSION_ID
canfar stats

Do not assume that headless work has priority over interactive work. Queue policy is deployment-owned. See batch processing.

What does canfar create print?

Human mode reports the accepted Session ID. JSON and YAML modes emit the raw list of returned IDs, suitable for a script:

canfar create headless IMAGE_NAME --output json -- python run.py

Put CLI options before the -- delimiter; everything after it is the command given to the container. A partial result contains the IDs accepted by the platform. An empty result indicates a creation or transport failure, not a queued Session.

Where should I put data?

  • /arc/home/<user>/ for personal persistent files;
  • /arc/projects/<project>/ for project files, when that path is available;
  • a configured VOSpace Service for persistent remote data; and
  • /scratch for temporary, high-speed staging only.

Use explicit IDENTIFIER:/path operands with canfar data. See Storage and Data transfers.

How do I use storage from Python?

The public API uses explicit identifiers:

from canfar.storage import filesystem, identifiers

print(identifiers())
remote = filesystem("IDENTIFIER")
try:
    print(remote.ls("/path"))
finally:
    remote.close()

local is the machine running the Python process. CANFAR does not register a dynamic vault:// or arc:// protocol and there is no storage.configure() call. See Filesystem and Python tools.

Can I use a GPU or request more resources?

Use the resource options supported by the current canfar create --help and the image's requirements. A larger fixed request can wait longer for matching capacity. Record the image and resource request with a reproducible workflow. Ask the platform operator about deployment-specific GPU availability.

How do I build or publish a Container Image?

Start with Container Images and the registry instructions for your project. Use a stable tag or digest and never store credentials in an image layer. A pushed image is usable only where the target Science Platform Server can pull it.

How do I get help?

Run the checks in Support, then email support@canfar.net with a Server Name, timestamp, exact error, and relevant Session diagnostics. Remove tokens, certificates, passwords, and private data before sharing logs.