Setup Shared Tool Infrastructure for internal LibreCores nodes

[oleg-nenashev]

Requirements:

• Availability via network (shared storage on NFS)
• Support of multiple tool versions with automatic version listing
• Nice2Have: Integration with Custom Tools (environment setup, etc.)
  • Still needs Pipeline integration :(

Open questions:

• Samba/NFS or Amazon-specific options?

[Houkime]

Is this still an issue? How does this translate onto dockerized infra?
Example:
You can have different slave Docker containers with different tools. Maybe even reconfigurable ones. Master node will run different jobs and maybe even job stages on different nodes. And maybe sometimes ask nodes (excluding ones that are user-donated) to reconfigure themselves. Custom tools/FPGAs available are reported by the nodes when they first call for master.
Why will you ever need NFS?

[oleg-nenashev]

We started from using containers with tools. Unfortunately, some containers in this case become too big, e.g. when you run FuseSoC. In this case you need to have all tools in a single container. In my test setup it's solved by a volume with tools, which are deployable via Modules. But it still needs to be documented better