Diffrence between execution and CAS shard worker about bazel-buildfarm HOT 3 CLOSED

1. Not sure what is missing from the docs on the difference between the two behaviors. A worker can be one of execute or CAS shard, or both. Action results are currently stored on the worker that executes the work if it is also a CAS shard, may be replicated to other workers that use the results as inputs, and are uploaded to CAS shard workers if the worker is execute-only.
2. Look in the example config.yml under capabilities: https://bazelbuild.github.io/bazel-buildfarm/docs/configuration/configuration/#capabilities. Workers register themselves uniquely to Workers_execute and Workers_storage hash sets in redis. There's no externally presented mechanism for distinguishing their type currently - what is your use case for this?
3. You can specify a bazel-remote (or any other grpc endpoint) as a secondary/terminal storage layer for a worker after the filesytem. It will read through that CAS to present content to clients seamlessly. See https://bazelbuild.github.io/bazel-buildfarm/docs/architecture/CASFileCache/#delegation and the grpc storage example in https://bazelbuild.github.io/bazel-buildfarm/docs/configuration/configuration/#worker-cas

from bazel-buildfarm.

Assuming the answers here satisfy the questions posed. Feel free to reopen if you need further clarification.

from bazel-buildfarm.

Sry for a long period inactive. That's really helpful, thx a lot for your great reply 🙏

from bazel-buildfarm.