adds quic receiver to shred-fetch-stage

[behzadnouri]

Problem

Working towards migrating turbine to QUIC.

Summary of Changes

Added QUIC receiver to shred-fetch-stage.

[t-nelson]

is this something that can wait for 1.16 to be cut? really need to keep branch diffs to a minimum for the time being

[behzadnouri]

is this something that can wait for 1.16 to be cut? really need to keep branch diffs to a minimum for the time being

I actually need this to be deployed to testnet asap,
so that we can test performance on an actual cluster and optimize where needed.

[behzadnouri]

Can I have this reviewed/merged?
I also have #31610 depending on this change.

Also these PRs only add a code-path which is inactive for all shreds. The goal is to incrementally rollout and monitor for small percent of shreds on testnet and debug/optimize based on the outcome.

[codecov]

Codecov Report

Merging #31576 (9cb1fc2) into master (8937fd9) will increase coverage by 0.0%.
The diff coverage is 94.1%.

@@           Coverage Diff           @@
##           master   #31576   +/-   ##
=======================================
  Coverage    81.8%    81.9%           
=======================================
  Files         763      763           
  Lines      207690   207735   +45     
=======================================
+ Hits       170067   170156   +89     
+ Misses      37623    37579   -44     

[t-nelson]

we teed this feature set up for discussion on the thursday priorities call. general consensus today was that holding quic turbine for 1.16 would fair better for 1.16 stabilization

[steviez]

Overall, the code looks good. I see merit to Trent's point about not wanting to introduce more changes between v1.14 and v1.16 given the already massive diff between the two. However, I also see merit to Behzad's point about wanting to get some basic plumbing in to build on top of to enable initial testing without having to deal with rebase hell later.

In order to try to meet in the middle, maybe something like a5f290a would be agreeable? That is, push the code in but not explicitly spawn the server by default for now.

[behzadnouri]

In order to try to meet in the middle, maybe something like a5f290a would be agreeable? That is, push the code in but not explicitly spawn the server by default for now.

We actually need to start testing QUIC on testnet for propagating a small percentage of slots.

[steviez]

With v1.16 branch cut, I believe we're good to go on these.

I see there have been some conversations in Discord and a SIMD, but assuming you're looking to push this in and test in parallel while those get hashed out ?