* Vote signing JSON RPC service
- barebone service that listens for RPC requests
* Daemon for vote signer service
* Add request APIs for JSON RPC
* Cleanup of cargo dependencies
* Fix compiler error
* Cleanup book
* Distinguish upstream from downstream validators
* Add BroadcastStage to Fullnode/Tpu diagrams
* First attempt to re-describe the runtime
* Reorg book
Push back details of the fullnode implementation
* Add db_window module for windowing functions from RocksDb
* Replace window with db_window functions in window_service
* Fix tests
* Make note of change in db_window
* Create RocksDb ledger in bin/fullnode
* Make db_ledger functions generic
* Add db_ledger to bin/replicator
First, talk about how a client interacts with Solana to do useful
things. Then describe how the fullnode you're talking to works and
why it's so very fast. Last, why that fullnode you don't trust
does what you asked it to anyway.
* Merge Leader and Validator diagrams
* New sdk-tools diagram
* Move terminology to just after introduction
* Purge use of LAMPORT as an acronym
* Add notes about persistent storage
An RPC client that fetches the signature status before the bank finishes
executing the corresponding Transaction should receive SignatureNotFound
instead of Confirmed
* Set drone address to always be the initial network entry point, so that even when leaders rotate the client can still find the drone
* Extract drone address as a separate argument to bench-tps
* Add drone port to client.sh instead of setting it in bench-tps
* Add drone entrypoint to scripts
* Fix build error
* Cluster Replicated Data Store
Separate the data storage and merge strategy from the network IO boundary.
Implement an eager push overlay for transporting recent messages.
Simulation shows fast convergence with 20k nodes.
That's supposed to be an ASCII format, but we're not making use
of it. We can switch back to that some day, but if we do, it shouldn't
be JSON-encoded.
* Add the solana-wallet documentation
There doesn't seem to be a way to publish bin docs to crates.io.
Until there is, we can include CLI documentation is the appendix
of the markdown book.
* A command to generate all the usage docs
Usage:
$ scripts/wallet-help.sh >> src/wallet.md
* Reorg TVU code to look like TVU diagram
And move channel creation into LedgerWriteStage so that it can
be used in the same was as all the other stages.
* Delete commented out code
* Stub out architecture documentation
* Add book HTML generation and book tests to CI
* Add heading
* Better table of contents
* Reference existing documentation
Move ASCII art from code comments into rendered SVG
* Attempt to fix CI
* Add lamport docs
And truncate lines to 80 characters
* Fix links
And reference shorter, newer description of PoH.
* Replace ASCII art with SVG
* Streamline for Pillbox
* Update path before optional install
* Use $CARGO_HOME instead of $HOME
* Delete code
Attempt to describe all data structures without code.
* Boot RPU from docs, add JsonRpcService
Also, use Rust naming conventions in the block diagrams to
minimize the jump from docs to code.
* Latest code uses tick_height
* Rename bob/ folder to art/
A home for any ASCII art
* Import JSON RPC API
* More mdbook docs
* Add Ncp
* Cleanup links
* Move pipelining description into fullnode description
* Move high-level transaction docs into top-level doc
* Delete unused files
add linked-list capability to accounts
change accounts from a linked list to a VecDeque
add checkpoint and rollback for lastids
add subscriber notifications for rollbacks
checkpoint transaction count, too
* Move finality computation into a service run from the banking stage, ComputeLeaderFinalityService
* Change last ids nth to tick height, remove separate tick height from bank
* Add first leader to genesis entries, consume in genesis.sh
* Set bootstrap leader in the bank on startup, remove instantiation of bootstrap leader from bin/fullnode
* Remove need to initialize bootstrap leader in leader_scheduler, now can be read from genesis entries
* Add separate interface new_with_leader() in mint for creating genesis leader entries
Deserialize operations are faster when done serially with the
MT banking stage and helps with performance improvement with
reduced thread context switches.
* Added tests to thin client to test VoteContract calls, fix VoteContract sizing errors
* Calculate upper bound on VoteProgram size at runtime, add test for serializing/deserializing a max sized VoteProgram state
* Add Vote Contract
* Move ownership of LeaderScheduler from Fullnode to the bank
* Modified ReplicateStage to consume leader information from bank
* Restart RPC Services in Leader To Validator Transition
* Make VoteContract Context Free
* Remove voting from ClusterInfo and Tpu
* Remove dependency on ActiveValidators in LeaderScheduler
* Switch VoteContract to have two steps 1) Register 2) Vote. Change thin client to create + register a voting account on fullnode startup
* Remove check in leader_to_validator transition for unique references to bank, b/c jsonrpc service and rpcpubsub hold references through jsonhttpserver
* Use IV to make unique identies
* Use hex! macro for hex literal and not string converted to u8 slice
* fix sha sampling to control init/end of sha state
* Move ledger write to its own stage
- Also, rename write_stage to leader_vote_stage, as write functionality
is moved to a different stage
* Address review comments
* Fix leader rotation test failure
* address review comments
Addresses the following problem
- Validators are not able to keep up with the leader
- The future blobs (outside of window) get dropped
- The validators won't process repair requests for these future blobs
* Add PoH height to process_ledger()
* Moved broadcast_stage Leader Scheduling logic to use Poh height instead of entry_height
* Moved LeaderScheduler logic to PoH in ReplicateStage
* Fix Leader scheduling tests to use PoH instead of entry height
* Change is_leader detection in repair() to use PoH instead of entry height
* Add tests to LeaderScheduler for new functionality
* fix Entry::new and genesis block PoH counts
* Moved LeaderScheduler to PoH ticks
* Cleanup to resolve PR comments
Budget now assumes the source account holds all tokens the program
should spend.
Note: the static guarantees implied by verify_plan() are meaningless
under the new contract engine. The bank no longer calls it. This
serves as a nice example of where comparing code coverage between
integration tests and unit tests would have shown us where a
change rendered unit tests meaningless.
Debits no longer need to be applied before credits. Instead, we
lock any accounts we'd debit and so error out on the second attempt
to lock the same account.
In the old bank (before the contract engine), Contract wasn't specific
to Budget. It provided the same service as what is now called
SystemProgram::Move, but without requiring a separate account.
- Testnet dashboard shows that channel pressure for write stage
is incrementing on every iteration of write.
- This change optimizes ledger writing by removing cloning of map
and reducing calls to flush
* Upload bench output as build artifacts
* Fix tags types
* Pull previous stats from metrics
* Change the default branch for comparison
* Fix formatting
* Fix build errors
* Address review comments
* Dedup some common code
* Add eval for channel info to find branch name
Generate tick entry ids and only register ticks as the last_id expected by the bank. Since the bank is MT, the in-flight pipeline of transactions cannot be close to the end of the queue or there is a high possibility that a starved thread will encode an expired last_id into the ledger. The banking_stage therefore uses a shorter age limit for encoded last_ids then the validators.
Bench client doesn't send transactions that are older then 30 seconds.
* Added LeaderScheduler module and tests
* plumbing for LeaderScheduler in Fullnode + tests. Add vote processing for active set to ReplicateStage and WriteStage
* Add LeaderScheduler plumbing for Tvu, window, and tests
* Fix bank and switch tests to use new LeaderScheduler
* move leader rotation check from window service to replicate stage
* Add replicate_stage leader rotation exit test
* removed leader scheduler from the window service and associated modules/tests
* Corrected is_leader calculation in repair() function in window.rs
* Integrate LeaderScheduler with write_stage for leader to validator transitions
* Integrated LeaderScheduler with BroadcastStage
* Removed gossip leader rotation from crdt
* Add multi validator, leader test
* Comments and cleanup
* Remove unneeded checks from broadcast stage
* Fix case where a validator/leader need to immediately transition on startup after reading ledger and seeing they are not in the correct role
* Set new leader in validator -> validator transitions
* Clean up for PR comments, refactor LeaderScheduler from process_entry/process_ledger_tail
* Cleaned out LeaderScheduler options, implemented LeaderScheduler strategy that only picks the bootstrap leader to support existing tests, drone/airdrops
* Ignore test_full_leader_validator_network test due to bug where the next leader in line fails to get the last entry before rotation (b/c it hasn't started up yet). Added a test test_dropped_handoff_recovery go track this bug
Michael Vines
carllin
Sagar Dhawan
Pankaj Garg
Rob Walker
Stephen Akridge
Greg Fitzgerald
Sathish
jackcmay
Anatoly Yakovenko
anatoly yakovenko
Tyera Eulberg
Tyera Eulberg