George Gensure

Can you please post a relevant experimental grpc remote log from a run with a failure (something with just the reported ByteStream::Write invocation would suffice) - I'm concerned that the...

Deleted previous recommendation. I believe this was caused by not reporting an uncompressed committed size, and experiencing a client restart. I've put up https://github.com/werkt/bazel-buildfarm/tree/zstd-committed-size to correct for the reporting violation.

Workers must have a FILESYSTEM type cas as its first element. Otherwise it will not be able to execute, or have function as a shard member

Currently, the only replication to a delegate CAS for FILESYSTEM is on expiration (think waterfall). Once a worker CAS is filled, in order to add new blobs, we spill over...

Cannot confirm. Missing single directory will be created by worker process. Path with missing directory component will be created with `Files.createDirectory` if everything can be created, and the application will...

The requirements for this *may* be that skipLoad is true

I believe this is no longer an issue, as we've been using bazel 7 for some time now. Closing.

I commented on the original PR for aacd1d77d8, and this thread seems dead as well, but I have done the same synchronization mechanism, available without modification of the jedis library...

@andrewrothstein can you review since push?

Reading this, it looks like a bazel client-sourced exception not on connection to buildfarm, but on cert load, which this project would have no ability to affect. Recommend you post...