* PVF: Remove `rayon` and some uses of `tokio`
1. We were using `rayon` to spawn a superfluous thread to do execution, so it was removed.
2. We were using `rayon` to set a threadpool-specific thread stack size, and AFAIK we couldn't do that with `tokio` (it's possible [per-runtime](https://docs.rs/tokio/latest/tokio/runtime/struct.Builder.html#method.thread_stack_size) but not per-thread). Since we want to remove `tokio` from the workers [anyway](https://github.com/paritytech/polkadot/issues/7117), I changed it to spawn threads with the `std::thread` API instead of `tokio`.[^1]
[^1]: NOTE: This PR does not totally remove the `tokio` dependency just yet.
3. Since `std::thread` API is not async, we could no longer `select!` on the threads as futures, so the `select!` was changed to a naive loop.
4. The order of thread selection was flipped to make (3) sound (see note in code).
I left some TODO's related to panics which I'm going to address soon as part of https://github.com/paritytech/polkadot/issues/7045.
* PVF: Vote invalid on panics in execution thread (after a retry)
Also make sure we kill the worker process on panic errors and internal errors to
potentially clear any error states independent of the candidate.
* Address a couple of TODOs
Addresses a couple of follow-up TODOs from
https://github.com/paritytech/polkadot/pull/7153.
* Add some documentation to implementer's guide
* Fix compile error
* Fix compile errors
* Fix compile error
* Update roadmap/implementers-guide/src/node/utility/candidate-validation.md
Co-authored-by: Andrei Sandu <54316454+sandreim@users.noreply.github.com>
* Address comments + couple other changes (see message)
- Measure the CPU time in the prepare thread, so the observed time is not
affected by any delays in joining on the thread.
- Measure the full CPU time in the execute thread.
* Implement proper thread synchronization
Use condvars i.e. `Arc::new((Mutex::new(true), Condvar::new()))` as per the std
docs.
Considered also using a condvar to signal the CPU thread to end, in place of an
mpsc channel. This was not done because `Condvar::wait_timeout_while` is
documented as being imprecise, and `mpsc::Receiver::recv_timeout` is not
documented as such. Also, we would need a separate condvar, to avoid this case:
the worker thread finishes its job, notifies the condvar, the CPU thread returns
first, and we join on it and not the worker thread. So it was simpler to leave
this part as is.
* Catch panics in threads so we always notify condvar
* Use `WaitOutcome` enum instead of bool condition variable
* Fix retry timeouts to depend on exec timeout kind
* Address review comments
* Make the API for condvars in workers nicer
* Add a doc
* Use condvar for memory stats thread
* Small refactor
* Enumerate internal validation errors in an enum
* Fix comment
* Add a log
* Fix test
* Update variant naming
* Address a missed TODO
---------
Co-authored-by: Andrei Sandu <54316454+sandreim@users.noreply.github.com>
* PVF: Don't dispute on missing artifact
A dispute should never be raised if the local cache doesn't provide a certain
artifact. You can not dispute based on this reason, as it is a local hardware
issue and not related to the candidate to check.
Design:
Currently we assume that if we prepared an artifact, it remains there on-disk
until we prune it, i.e. we never check again if it's still there.
We can change it so that instead of artifact-not-found triggering a dispute, we
retry once (like we do for AmbiguousWorkerDeath, except we don't dispute if it
still doesn't work). And when enqueuing an execute job, we check for the
artifact on-disk, and start preparation if not found.
Changes:
- [x] Integration test (should fail without the following changes)
- [x] Check if artifact exists when executing, prepare if not
- [x] Return an internal error when file is missing
- [x] Retry once on internal errors
- [x] Document design (update impl guide)
* Add some context to wasm error message (it is quite long)
* Fix impl guide
* Add check for missing/inaccessible file
* Add comment referencing Substrate issue
* Add test for retrying internal errors
---------
Co-authored-by: parity-processbot <>
* Happy New Year!
* Remove year entierly
Co-authored-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
* Remove years from copyright notice in the entire repo
---------
Co-authored-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
* Use a `BoundedVec` in `ValidationResult`
> Use a `BoundedVec` for `upward_messages` and `horizontal_messages` in order to
> limit the number of individual messages/memory allocations right at decoding
> time. The reason for this is that the `ValidationResult` may contain a code
> upgrade (including a full PVF binary), so the total size limit can't be set
> too low and this limit will still allow several millions of upward messages,
> which will (due to the memory allocator overhead) already have a
> non-negligible memory footprint in decoded form.
* List all fields when hashing so we don't miss one
* Define types for `BoundedVec`s of messages
* Fix test compile errors
* Depend on `bounded-collections` 0.1.4 (fixes allocation issue)
* Fix compilation issue
* Derive `Hash` instead of manual `impl`
* Avoid use of unwrap
* Re-apply changes without Diener, rebase to the lastest master
* Cache pruning
* Bit-pack InstantiationStrategy
* Move ExecutorParams version inside the structure itself
* Rework runtime API and executor parameters storage
* Pass executor parameters through backing subsystem
* Update Cargo.lock
* Introduce `ExecutorParams` to approval voting subsys
* Introduce `ExecutorParams` to dispute coordinator
* `cargo fmt`
* Simplify requests from backing subsys
* Fix tests
* Replace manual config cloning with `.clone()`
* Move constants to module
* Parametrize executor performing PVF pre-check
* Fix Malus
* Fix test runtime
* Introduce session executor params as a constant defined by session info
pallet
* Use Parity SCALE codec instead of hand-crafted binary encoding
* Get rid of constants; Add docs
* Get rid of constants
* Minor typo
* Fix Malus after rebase
* `cargo fmt`
* Use transparent SCALE encoding instead of explicit
* Clean up
* Get rid of relay parent to session index mapping
* Join environment type and version in a single enum element
* Use default execution parameters if running an old runtime
* `unwrap()` -> `expect()`
* Correct API version
* Constants are back in town
* Use constants for execution environment types
* Artifact separation, first try
* Get rid of explicit version
* PVF execution queue worker separation
* Worker handshake
* Global renaming
* Minor fixes resolving discussions
* Two-stage requesting of executor params to make use of runtime API cache
* Proper error handling in pvf-checker
* Executor params storage bootstrapping
* Propagate migration to v3 network runtimes
* Fix storage versioning
* Ensure `ExecutorParams` serialization determinism; Add comments
* Rename constants to make things a bit more deterministic
Get rid of stale code
* Tidy up a structure of active PVFs
* Minor formatting
* Fix comment
* Add try-runtime hooks
* Add storage version write on upgrade
Co-authored-by: Andronik <write@reusable.software>
* Add pre- and post-upgrade assertions
* Require to specify environment type; Remove redundant `impl`s
* Add `ExecutorParamHash` creation from `H256`
* Fix candidate validation subsys tests
* Return splittable error from executor params request fn
* Revert "Return splittable error from executor params request fn"
This reverts commit a0b274177d8bb2f6e13c066741892ecd2e72a456.
* Decompose approval voting metrics
* Use more relevant errors
* Minor formatting fix
* Assert a valid environment type instead of checking
* Fix `try-runtime` hooks
* After-merge fixes
* Add migration logs
* Remove dead code
* Fix tests
* Fix tests
* Back to the strongly typed implementation
* Promote strong types to executor interface
* Remove stale comment
* Move executor params to `SessionInfo`: primitives and runtime
* Move executor params to `SessionInfo`: node
* Try to bump primitives and API version
* Get rid of `MallocSizeOf`
* Bump target API version to v4
* Make use of session index already in place
* Back to v3
* Fix all the tests
* Add migrations to all the runtimes
* Make use of existing `SessionInfo` in approval voting subsys
* Rename `TARGET` -> `LOG_TARGET`
* Bump all the primitives to v3
* Fix Rococo ParachainHost API version
* Use `RollingSessionWindow` to acquire `ExecutorParams` in disputes
* Fix nits from discussions; add comments
* Re-evaluate queue logic
* Rework job assignment in execution queue
* Add documentation
* Use `RuntimeInfo` to obtain `SessionInfo` (with blackjack and caching)
* Couple `Pvf` with `ExecutorParams` wherever possible
* Put members of `PvfWithExecutorParams` under `Arc` for cheap cloning
* Fix comment
* Fix CI tests
* Fix clippy warnings
* Address nits from discussions
* Add a placeholder for raw data
* Fix non exhaustive match
* Remove redundant reexports and fix imports
* Keep only necessary semantic features, as discussed
* Rework `RuntimeInfo` to support mock implementation for tests
* Remove unneeded bound
* `cargo fmt`
* Revert "Remove unneeded bound"
This reverts commit 932463f26b00ce290e1e61848eb9328632ef8a61.
* Fix PVF host tests
* Fix PVF checker tests
* Fix overseer declarations
* Simplify tests
* `MAX_KEEP_WAITING` timeout based on `BACKGING_EXECUTION_TIMEOUT`
* Add a unit test for varying executor parameters
* Minor fixes from discussions
* Add prechecking max. memory parameter (see paritytech/srlabs_findings#110)
* Fix and improve a test
* Remove `ExecutionEnvironment` and `RawData`
* New primitives versioning in parachain host API
* `disputes()` implementation for Kusama and Polkadot
* Move `ExecutorParams` from `vstaging` to stable primitives
* Move disputes from `vstaging` to stable implementation
* Fix `try-runtime`
* Fixes after merge
* Move `ExecutorParams` to the bottom of `SessionInfo`
* Revert "Move executor params to `SessionInfo`: primitives and runtime"
This reverts commit dfcfb85fefd1c5be6c8a8f72dc09fd1809cfa9ce.
* Always use fresh activated live hash in pvf precheck
(re-apply 34b09a4c20de17e7926ed942cd0d657d18f743fa)
* Fixing tests (broken commit)
* Fix candidate validation tests
* Fix PVF host test
* Minor fixes
* Address discussions
* Restore migration
* Fix `use` to only include what is needed instead of `*`
* Add comment to never touch `DEFAULT_CONFIG`
* Update migration to set default `ExecutorParams` for `dispute_period`
sessions back
* Use `earliest_stored_session` instead of calculations
* Nit
* Add logs
* Treat any runtime error as `NotSupported` again
* Always return default executor params if not available
* Revert "Always return default executor params if not available"
This reverts commit b58ac4482ef444c67a9852d5776550d08e312f30.
* Add paritytech/substrate#9997 workaround
* `cargo fmt`
* Remove migration (again!)
* Bump executor params to API v4 (backport from #6698)
---------
Co-authored-by: Andronik <write@reusable.software>
* Refactor PVF preparation memory stats
The original purpose of this change was to gate metrics that are unsupported by
some systems behind conditional compilation directives (#[cfg]); see
https://github.com/paritytech/polkadot/pull/6675#discussion_r1099996209.
Then I started doing some random cleanups and simplifications and got a bit
carried away. 🙈 The code should be overall tidier than before.
Changes:
- Don't register unsupported metrics (e.g. `max_rss` on non-Linux systems)
- Introduce `PrepareStats` struct as an abstraction over the `Ok` values of
`PrepareResult`. It is cleaner, and can be easily modified in the future.
- Other small changes
* Minor fixes to comments
* Fix compile errors
* Try to fix some Linux errors
* Mep
* Fix candidate-validation tests
* Update docstring
* pre-checking: Reject failed PVFs
* paras: immediately reject any PVF that cannot reach a supermajority
* Make the `quorum` reject condition a bit more clear semantically
* Add comment
* Update implementer's guide
* Update a link
Not related to the rest of the PR, but I randomly noticed and fixed this.
* Update runtime/parachains/src/paras/tests.rs
Co-authored-by: s0me0ne-unkn0wn <48632512+s0me0ne-unkn0wn@users.noreply.github.com>
* Remove unneeded loop
* Log PVF retries using `info!`
* Change retry logs to `warn!` and add preparation failure log
* Log PVF execution failure
* Clarify why we reject failed PVFs
* Fix PVF reject runtime benchmarks
Co-authored-by: s0me0ne-unkn0wn <48632512+s0me0ne-unkn0wn@users.noreply.github.com>
* Replace async-std with tokio in PVF subsystem
* Rework workers to use `select!` instead of a mutex
The improvement in code readability is more important than the thread overhead.
* Remove unnecessary `fuse`
* Add explanation for `expect()`
* Update node/core/pvf/src/worker_common.rs
Co-authored-by: Bastian Köcher <info@kchr.de>
* Update node/core/pvf/src/worker_common.rs
Co-authored-by: Bastian Köcher <info@kchr.de>
* Address some review comments
* Shutdown tokio runtime
* Run cargo fmt
* Add a small note about retries
* Fix up merge
* Rework `cpu_time_monitor_loop` to return when other thread finishes
* Add error string to PrepareError::IoErr variant
* Log when artifacts fail to prepare
* Fix `cpu_time_monitor_loop`; fix test
* Fix text
* Fix a couple of potential minor data races.
First data race was due to logging in the CPU monitor thread even if the
job (other thread) finished. It can technically finish before or after the log.
Maybe best would be to move this log to the `select!`s, where we are guaranteed
to have chosen the timed-out branch, although there would be a bit of
duplication.
Also, it was possible for this thread to complete before we executed
`finished_tx.send` in the other thread, which would trigger an error as the
receiver has already been dropped. And right now, such a spurious error from
`send` would be returned even if the job otherwise succeeded.
* Update Cargo.lock
Co-authored-by: Bastian Köcher <info@kchr.de>
* PVF preparation: do not conflate errors
+ Adds some more granularity to the prepare errors.
+ Better distinguish whether errors occur on the host side or the worker.
+ Do not kill the worker if the error happened on the host side.
+ Do not retry preparation if the error was `Panic`.
+ Removes unnecessary indirection with `Selected` type.
* Add missing docs, resolve TODOs
* Address review comments and remove TODOs
* Fix error in CI
* Undo unnecessary change
* Update couple of comments
* Don't return error for stream shutdown
* Update node/core/pvf/src/worker_common.rs
* Put in skeleton logic for CPU-time-preparation
Still needed:
- Flesh out logic
- Refactor some spots
- Tests
* Continue filling in logic for prepare worker CPU time changes
* Fix compiler errors
* Update lenience factor
* Fix some clippy lints for PVF module
* Fix compilation errors
* Address some review comments
* Add logging
* Add another log
* Address some review comments; change Mutex to AtomicBool
* Refactor handling response bytes
* Add CPU clock timeout logic for execute jobs
* Properly handle AtomicBool flag
* Use `Ordering::Relaxed`
* Refactor thread coordination logic
* Fix bug
* Add some timing information to execute tests
* Add section about the mitigation to the IG
* minor: Change more `Ordering`s to `Relaxed`
* candidate-validation: Fix build errors
* Add clippy config and remove .cargo from gitignore
* first fixes
* Clippyfied
* Add clippy CI job
* comment out rusty-cachier
* minor
* fix ci
* remove DAG from check-dependent-project
* add DAG to clippy
Co-authored-by: alvicsam <alvicsam@gmail.com>
* Fix a couple of typos
* Retry failed PVF execution
PVF execution that fails due to AmbiguousWorkerDeath should be retried once.
This should reduce the occurrence of failures due to transient conditions.
Closes#6195
* Address a couple of nits
* Write tests; refactor (add `validate_candidate_with_retry`)
* Update node/core/candidate-validation/src/lib.rs
Co-authored-by: Andronik <write@reusable.software>
Co-authored-by: eskimor <eskimor@users.noreply.github.com>
Co-authored-by: Andronik <write@reusable.software>
* split metrics from collation generation
* move metrics to separate file out of backing
* split bitfield signing metrics
* split candidate validation metrics
* split chain api metrics
* split metrics from runtime API
* util is not used in backed metrics mod
* fmt
* missing types
* sure
* remove v0 primitives from polkadot-primitives
* first pass: remove v0
* fix fallout in erasure-coding
* remove v1 primitives, consolidate to v2
* the great import update
* update runtime_api_impl_v1 to v2 as well
* guide: add `Version` request for runtime API
* add version query to runtime API
* reintroduce OldV1SessionInfo in a limited way
* Companion PR for removing Prometheus metrics prefix
* Was missing some metrics
* Fix missing renames
* Fix test
* Fixes
* Update test
* Update Substrate
* Second time
* remove prefix from intergration test for zombienet
* update zombienet image
* Update Substrate
Co-authored-by: Bastian Köcher <info@kchr.de>
Co-authored-by: Javier Viola <pepoviola@gmail.com>
* remove Default from CandidateHash
* Apply suggestions from code review
Co-authored-by: Andronik Ordian <write@reusable.software>
* chore: fmt
* remove backed candidate default
* Partial migration away from CandidateReceipt::default
* Remove more CandidateReceipt defaults
* fmt
* Mostly remove CommittedCandidateReceipt default usage
* Remove CommittedCandidateReceipt
* Remove more Defaults from polakdot primitives v1 + fmt
* Remove more Default from polkadot primites v1
* WIP trying to get overseer example + tests to compile
* feat: add primitives test helpers
* reduce deps of helper
* update primitive helpers
* make candidate validation compile
* fixup cargo lock
* make av-store compile
* fixup disputes coordinator tests
* test: fixup backing
* test: fixup approval voting
* fixup bitfield signing
* test: fixup runtime-api
* test: fixup availability dist
* foxi[ pverseer test]
* remove some Defaults, remove bounds from `dummy`
All `fn dummy` in primitives need to be removed anyways.
This aids in the transition.
* it's a test helper, so always use std
* test: fixup parachains runtime tests
Excluding benches.
* fix keyring
* fix paras runtime properly, no more default
* Remove fn dummy() usage from approval voting
* Move TestCandidateBuilder out of av store to test helpers
* Make candidate validation tests pass
* Make most dispute coirdinator tests pass
* Make provisioner tests work
* Make availability recovery tests work with test helpers
* Update polkadot-collator-protocol tests
* Update statement distribution tests
* Update polkadot overseer examples and tests
* Derive default for validation code so we don't break unrelated things
* Make para runtime test pass (no bench)
* Some more work
* chore: cargo fmt
* cargo fix
* avoid some Default::default
* fixup dispute coordinator test
* remove unused crate deps
* remove Default::default wherever possible, replace by dummy_* for the most part
* chore: cargo fmt
* Remove some warnings
* Remove CommittedCandidateReceipt dummy
* Remove CandidateReceipt dummy
* Remove CandidateDescriptor dummy
* Remove commented out code
* Fix para runtime tests
* chore: nightly
* Some updates to the builder
* Dynamically adjust mock head data size
* Make dispute cooridinator tests work
* Fix test candidate_backing_reorders_votes work
* +nightly-2021-10-29 fmt
* Spelling and remove a default use in builder
* Various clean up
* More small updates
* fmt
* More small updates
* Doc comments for test helpers
* cargo run --quiet --release --features=runtime-benchmarks -- benchmark --chain=kusama-dev --steps=50 --repeat=20 --pallet=runtime_parachains::paras_inherent --extrinsic=* --execution=wasm --wasm-execution=compiled --heap-pages=4096 --header=./file_header.txt --output=./runtime/kusama/src/weights/runtime_parachains_paras_inherent.rs
* cargo run --quiet --release --features=runtime-benchmarks -- benchmark --chain=polkadot-dev --steps=50 --repeat=20 --pallet=runtime_parachains::paras_inherent --extrinsic=* --execution=wasm --wasm-execution=compiled --heap-pages=4096 --header=./file_header.txt --output=./runtime/polkadot/src/weights/runtime_parachains_paras_inherent.rs
* Update lib.rs
* review comments
* fix warnings
* fix test by using correct candidate receipt relay parent
Co-authored-by: Andronik Ordian <write@reusable.software>
Co-authored-by: emostov <32168567+emostov@users.noreply.github.com>
Co-authored-by: Parity Bot <admin@parity.io>
Co-authored-by: Gavin Wood <gavin@parity.io>
Closes https://github.com/paritytech/polkadot/issues/4293
This PR changes the way how we treat a certain subset of PVF preparation
errors. Specifically, now only the deterministic errors are treated as
invalid candidates. That is, the errors that are easily
attributable to either the the PVF contents or the wasmtime code, but
not e.g. I/O errors that could be triggered by the OS (insufficient
memory, disk failure, too much load, etc). The latter are treated as
internal errors and thus do not trigger the disputes.
* pvf: make execution timeout configurable
* guide: add timeouts to candidate validation params
* add timeouts to candidate validation messages
* fmt
* port backing to use the backing pvf timeout
* port approval-voting to use the execution timeout
* port dispute participation to use the correct timeout
* fmt
* address grumbles & test failure
* feat/overseer: introduce closure init
Enables removal of the connected/disconnected overseer state.
* feat/overseer: allow replacement logic to access the original
Allows to re-use init-once types, which would otherwise error.
* feat/overseer: introduce external connector
Preparation for removal of `AllSubsystems`
which is another prerequisite for removing
the connect/disconnect state.
* fix/test: replace needs closure
* fixup
* simplify
* mea culpa
* all-subsystems-gen test
* Factor out runtime module into utils.
* Add maybe_authority information to `PeerConnected` event.
We already gather this information in authority discovery, so we might
as well share it with others.
This opens up an easy path to trigger validators differently from normal
nodes, e.g. for prioritization. This change has become more important
now, that we just connect to all validators and therefore just have a
long peer list without any information about those nodes.
* Test fix.
* Implement PVF validation host
* WIP: Diener
* Increase the alloted compilation time
* Add more comments
* Minor clean up
* Apply suggestions from code review
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* Fix pruning artifact removal
* Fix formatting and newlines
* Fix the thread pool
* Update node/core/pvf/src/executor_intf.rs
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* Remove redundant test declaration
* Don't convert the path into an intermediate string
* Try to workaround the test failure
* Use the puppet_worker trick again
* Fix a blip
* Move `ensure_wasmtime_version` under the tests mod
* Add a macro for puppet_workers
* fix build for not real-overseer
* Rename the puppet worker for adder collator
* play it safe with the name of adder puppet worker
* Typo: triggered
* Add more comments
* Do not kill exec worker on every error
* Plumb Duration for timeouts
* typo: critical
* Add proofs
* Clean unused imports
* Revert "WIP: Diener"
This reverts commit b9f54e513366c7a6dfdd117ac19fbdc46b900b4d.
* Sync version of wasmtime
* Update cargo.lock
* Update Substrate
* Merge fixes still
* Update wasmtime version in test
* bastifmt
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* Squash spaces
* Trailing new line for testing.rs
* Remove controversial code
* comment about biasing
* Fix suggestion
* Add comments
* make it more clear why unwrap_err
* tmpfile retry
* proper proofs for claim_idle
* Remove mutex from ValidationHost
* Add some more logging
* Extract exec timeout into a constant
* Add some clarifying logging
* Use blake2_256
* Clean up the merge
Specifically the leftovers after removing real-overseer
* Update parachain/test-parachains/adder/collator/Cargo.toml
Co-authored-by: Andronik Ordian <write@reusable.software>
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
Co-authored-by: Andronik Ordian <write@reusable.software>
* use compressed blob in candidate-validation
* add some tests for compressed code blobs
* remove CompressedPoV and apply compression in collation-generation
* decompress BlockData before executing
* don't produce oversized collations
* add test for PoV decompression failure
* fix tests and clean up
* fix test
* address review and fix CI
* take this )
Marcin S