Problems with High-Performance Computing and Canonical Snapcraft. Any Advice?

Hi everyone,

I’m currently experiencing some challenges with my high-performance computing setup and a Canonical product. As someone relatively new to managing both hardware and software in this capacity, I could really use some advice from those more experienced.

Here are the specifics of my setup:

  • High-Performance Computing Setup: I have a custom-built cluster with multiple high-end CPUs, GPUs, ample RAM, and SSD storage capacity.
  • Canonical Product: Snapcraft

The Issues:

  1. Performance Challenges: Applications packaged with Snapcraft are running slower than expected on my high-performance computing setup, particularly during intensive tasks and concurrent operations. I’m uncertain whether this is due to configuration settings, infrastructure limitations, or other factors.
  2. Installation and Configuration Difficulties: During the setup of Snapcraft on my computing environment, I encountered several hurdles. While I managed to complete the installation, I suspect there may be lingering configuration issues affecting performance.
  3. Integration and Compatibility Concerns: I’m facing compatibility issues with certain libraries and dependencies within Snapcraft packages, leading to failed builds and incomplete functionalities.

Steps Taken So Far:

  • I’ve ensured that my hardware setup meets or exceeds the recommended specifications for Snapcraft.
  • I’ve followed the installation and setup documentation for Snapcraft meticulously.
  • I’ve reviewed and adjusted server configurations and resource allocations to optimize performance.

Despite my efforts, the challenges persist. Have any of you encountered similar issues with Snapcraft in high-performance computing environments? Are there specific configurations, optimizations, or troubleshooting techniques you’ve found effective? Any advice, insights, or recommended resources would be greatly appreciated.

Thank you in advance for your help!

Perhaps @nuccitheboss can help here, he leads the HPC community of Ubuntu…

Hi there @judywatsonn :wave:

Thank you for posting here! Couple of questions for you:

  1. How did you set up snapcraft and where are you running it? Is snapcraft installed on your local machine or is it on your cluster?
  2. Are your snapcraft.yaml recipes public or are they located somewhere I can view them? I’d be interested in checking them out.
  3. Which libraries and dependencies are you having issues with? We’ve been refining our practice with producing HPC-related snaps in the Ubuntu HPC community team, so perhaps we’ve encountered similar issues or have some ideas with where performance challenges are.

These questions will help me better understand what the snap environment our your cluster looks like :smiley:

cc @ogra thank you for summoning me :mage::magic_wand::sparkles:

2 Likes

Remember to check that your snaped applications doesn’t hit DENY by apparmor which causes them to revert to “compatibility mode” effectively running slower than on optimized mode. This often involves making sure access or availability to specialized libraries and custom configurations.

Knowing your application and how it interacts with the execution environment is MONUMENTAL if running an application in a snap.

That said, if you get it right it will be solid.

2 Likes