The `NeighborPacketWorker` within `client/finality-grandpa` does two
things:
1. It receives neighbor packets from components within
`client/finality-grandpa`, sends them down to the `GossipEngine` in
order for neighboring nodes to receive.
2. It periodically sends out the most recent neighbor packet to the
`GossipEngine`.
In order to send out packets it had a clone to a `GossipEgine` within
an atomic reference counter and a mutex. The `NeighborPacketWorker` was
then spawned onto its own asynchronous task.
Instead of running in its own task, this patch reintegrates the
`NeighborPacketWorker` into the main `client/finality-grandpa` task not
requiring the `NeighborPacketWorker` to own a clone of the
`GossipEngine`.
The greater picture
This is a tiny change within a greater refactoring. The overall goal is
to **simplify** how finality-grandpa interacts with the network and to
**reduce** the amount of **unbounded channels** within the logic.
Why no unbounded channels: Bounding channels is needed for backpressure
and proper scheduling. With unbounded channels there is no way of
telling the producer side to slow down for the consumer side to catch
up. Rephrased, there is no way for the scheduler to know when to favour
the consumer task over the producer task on a crowded channel and the
other way round for an empty channel.
Reducing the amount of shared ownership simplifies the logic and enables
one to use async-await syntax-suggar, given that one does not need to
hold a lock across poll invocations. Using async-await enables one to
use bounded channels without complex logic.
* Implement runtime version checks in `set_code`
Check that the new runtime code given to `set_code` fullfills some
requirements:
- `spec_name` matches
- `spec_version` does not decreases
- `impl_version` does not decreases
- Either `spec_version` and `impl_version` increase
* Make tests almost work
* Some fixes after master merge
* Fix tests
* Add missed file
* Make depedency check happy?
* Remove leftover `sc-executor`
* AHHHHH
* Reset debug stuff
* Remove some 'static
* More 'static
* Some docs
* Update `Cargo.lock`
* Make Proposer instantiation potentially async.
* fix node-service test
* fix basic-authority doc-test
* only block once on futures in test
* use async/await
* reset a couple of versions down from v0.2.0
* Unstablise browser-utils (very new) and grafana-data-source (going to be replaced with prometheus sometime)
* unmark a bunch more client crates for stabilization
* fix versions in Cargo.tomls
* Downgrade network-test
* Set frame-metadata version to `10.0.0`
* Hide and documented storage generator as internal api
* Downgrade `sp-externalities`
* Downgrade `sc-cli`
* Downgrade sc-executor et al to 0.8.0
* Downgrade sp-state-machine to 0.8.0
* Downgrade pallet-contracts et al to 0.8.0
* Downgrade sp-sandbox to 0.8.0
* downgrade pallet-evm to 0.8.0
* Downgrade pallet-staking to 0.8.0
2.0 should be implementation of lazy payout alongside all the fixes of
current module
* Downgrade node-transaction-factory to 0.8.0
* frame pallets are 2.0.0
Co-authored-by: Ashley <ashley.ruglys@gmail.com>
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
Co-authored-by: Sergei Pepyakin <s.pepyakin@gmail.com>
Co-authored-by: Wei Tang <accounts@that.world>
Co-authored-by: thiolliere <gui.thiolliere@gmail.com>
Co-authored-by: Marcio Diaz <marcio@parity.io>
* Switch service to futures03
* Fix tests
* Fix service test and cli
* Re-add Executor trait to SpawnTaskHandle
* Fix node-service
* Update babe
* Fix browser node
* Update aura
* Revert back to tokio-executor to fix runtime panic
* Add todo item
* Fix service tests again
* Timeout test futures
* Fix tests
* nits
* Fix service test
* Remove zstd patch
* Re-add futures01 to aura and babe tests as a dev-dep
* Change failing test to tee
* Fix node
* Upgrade tokio
* fix society
* Start switching grandpa to stable futures
* Revert "Start switching grandpa to stable futures"
This reverts commit 9c1976346237637effc07c13f7d0403daf5e71cf.
* Fix utils
* Revert substrate service test
* Revert gitlab
Co-authored-by: thiolliere <gui.thiolliere@gmail.com>
Instead of having these weird implementation of `BuildStorage for
&ChainSpec` we should just take the `genesis_storage` by ref. The
`BuildStorage` trait changed some time ago to take a self ref anyway,
instead of a self value.
Also fixes warnings in frame-staking
* fork-tree: rebalance fork tree by maximum branch height
* fork-tree: add docs for tree rebalancing
* fork-tree: add test
* babe: rebalance epoch changes tree on deserialization
* fork-tree: fix doc
* fork-tree: fix test
* Extend `Proposer` to optionally generate a proof of the proposal
* Something
* Refactor sr-api to not depend on client anymore
* Fix benches
* Apply suggestions from code review
Co-Authored-By: Tomasz Drwięga <tomusdrw@users.noreply.github.com>
* Apply suggestions from code review
* Introduce new `into_storage_changes` function
* Switch to runtime api for `execute_block` and don't require `H256`
anywhere in the code
* Put the `StorageChanges` into the `Proposal`
* Move the runtime api error to its own trait
* Adds `StorageTransactionCache` to the runtime api
This requires that we add `type NodeBlock = ` to the
`impl_runtime_apis!` macro to work around some bugs in rustc :(
* Remove `type NodeBlock` and switch to a "better" hack
* Start using the transaction cache from the runtime api
* Make it compile
* Move `InMemory` to its own file
* Make all tests work again
* Return block, storage_changes and proof from Blockbuilder::bake()
* Make sure that we use/set `storage_changes` when possible
* Add test
* Fix deadlock
* Remove accidentally added folders
* Introduce `RecordProof` as argument type to be more explicit
* Update client/src/client.rs
Co-Authored-By: Tomasz Drwięga <tomusdrw@users.noreply.github.com>
* Update primitives/state-machine/src/ext.rs
Co-Authored-By: Tomasz Drwięga <tomusdrw@users.noreply.github.com>
* Integrates review feedback
* Remove `unsafe` usage
* Update client/block-builder/src/lib.rs
Co-Authored-By: Benjamin Kampmann <ben@gnunicorn.org>
* Update client/src/call_executor.rs
* Bump versions
Co-authored-by: Tomasz Drwięga <tomusdrw@users.noreply.github.com>
Co-authored-by: Benjamin Kampmann <ben.kampmann@googlemail.com>
* Remove usage of sc_network::Context trait
* Remove Context::send_consensus
* Pass &mut dyn Network instead of &dyn Network
* Move Validator traits and related to separate module
* Entirely update substrate-telemetry to futures 0.3
* Add a Closed error
* Update to libp2p 0.14
* More work
* More work
* More work
* More work
* Fix warnings
* Remove unwrap()
* Work on tests fixing
* Fix network tests
* Fix external network tests
* Update libp2p and restore Yamux in discovery test
* Ignore DNS if initializatio nfails
* Restore variables ordering
* Forgot browser-utils
* Fix downfall after merge
* Fix tests
Previously one would create a sender and receiver channel pair, pass the
sender to the `build_network_future` through the service builder and
funnel network events returned from polling the network service into the
sender to be consumed by the authority discovery module owning the
receiver.
With recent changes it is now possible to register an `event_stream`
with the network service directly, thus one does not need to make the
detour through the `build_network_future`.
* client/authority-discovery: Limit number of connections to authorities
Instead of connecting to all sentry nodes of all authorities, with this
patch the authority discovery module does the following:
- Choose one sentry node per authority at random.
- Choose MAX_NUM_AUTHORITY_CONN out of the above at random.
The module uses randomness to prevent hot spots, e.g. all nodes trying
to connect to a single node. If the authority discovery module would
choose the nodes to connect to at random on each new address that it
learns of, the node would go through a lot of connection churn. Instead
it creates a random seed at start up and uses this seed for its RNG on
each update cycle.
* client/authority-discovery: Extract address cache into own module
* client/authority-discovery/src/addr_cache: Add basic unit tests
* client/authority-discovery: Replace unwrap with expect on [u8] cmp
* .maintain/sentry-node/docker-compose.yml: Prefix endpoint flags
* client/authority-discovery/src/addr_cache: Use sort_unstable and cmp
* client/authority-discovery: Use BTreeMap in addr_cache for sorted iter
To reduce connection churn it is preferrable to have `get_subset` of the
`addr_cache` to return the same result on repeated calls. `get_subset`
iterates a map. To make the process of iteration deterministic, use a
`BTreeMap` instead of a `HashMap`.
Max Inden