Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Parallel insertion of dirty store keys during clean #27058

Merged
merged 2 commits into from
Aug 20, 2022
Merged

Parallel insertion of dirty store keys during clean #27058

merged 2 commits into from
Aug 20, 2022

Conversation

bw-solana
Copy link
Contributor

Problem

While analyzing startup time, identified clean_accounts as a function that gets called several times, especially while shrinking slots. Execution time varies and depends heavily on key count, but takes on the order of seconds. Dirty store processing itself (part of construct_candidate_clean_keys) was taking multiple seconds. It also takes a couple of seconds to return from construct_candidate_clean_keys

Summary of Changes

  1. Parallelize dirty store key insertion <-- saves a few seconds per call
  2. "inline" construct_candidate_clean_keys into clean_accounts <-- saves a couple seconds per call

@bw-solana
Copy link
Contributor Author

bw-solana commented Aug 10, 2022 •
edited
Loading

Some A vs. B performance comparison data:
image

image

Baseline was taking ~7 seconds for dirty store processing and 10 seconds for overall clean. This code was taking ~800ms for dirty store processing and 4 seconds for overall clean.

Baseline master hash is e3a8d2ecdd358d836b14ebf80150492b1fb3b4d6 and data was collected during startup connecting a node to MNB.

@bw-solana bw-solana marked this pull request as ready for review August 11, 2022 01:42
@bw-solana bw-solana force-pushed the clean_accounts_optimization branch 2 times, most recently from 35b22a3 to b41d8e9 Compare August 17, 2022 17:41
jeffwashington
jeffwashington reviewed Aug 18, 2022
View reviewed changes
runtime/src/accounts_db.rs
let mut oldest_dirty_slot = max_slot.saturating_add(1);
dirty_store_chunk.iter().for_each(|(slot, store)| {
oldest_dirty_slot = oldest_dirty_slot.min(*slot);
store.accounts.account_iter().for_each(|account| {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes. This could be a page fault to iterate the append vec on disk. Makes sense we may want to request to page in multiple files in parallel as opposed to sequentially.

@jeffwashington jeffwashington self-requested a review August 18, 2022 16:48
jeffwashington
jeffwashington previously approved these changes Aug 18, 2022
View reviewed changes
Copy link
Contributor

@jeffwashington jeffwashington left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm.
rename pr to the part you're actually doing here.
Get rid of cargo.lock. rebasing will probably fix this.

@bw-solana bw-solana force-pushed the clean_accounts_optimization branch from b41d8e9 to 29ec1c7 Compare August 19, 2022 17:07
@mergify mergify bot dismissed jeffwashington’s stale review August 19, 2022 17:07

Pull request has been modified.

@bw-solana bw-solana changed the title Inline and parallelize construct_candidate_clean_keys Parallel insertion of dirty store keys during clean Aug 19, 2022
@bw-solana bw-solana merged commit 3c786ba into solana-labs:master Aug 20, 2022
@bw-solana bw-solana deleted the clean_accounts_optimization branch August 20, 2022 13:52
HaoranYi pushed a commit to HaoranYi/solana that referenced this pull request Aug 21, 2022
@bw-solana
Parallel insertion of dirty store keys during clean (solana-labs#27058)
4b4af89 
parallelize dirty store key insertion
HaoranYi added a commit that referenced this pull request Aug 22, 2022
@HaoranYi @jbiseda
@dependabot @brooksprumo @behzadnouri @anthontaylor @aschonfeld @apfitzge @jeffwashington @bw-solana @mvines @nickfrosty @jacobcreech @joncinque @yihau @jstarry @yhchiang-sol @leonardkulms @willhickey @tao-stones @xiangzhu70 @Lichtso
Refactor epoch reward 4 (#27261)
389bedd 
* refactor: extract store_stake_accounts fn

* refactor: extract store_vote_account fn

* refactor: extract reward history update fn

* remove avg point value from pay_valiator fn. not used

* clippy: slice

* clippy: slice

* remove abort() from test-validator (#27124)

* chore: bump bytes from 1.1.0 to 1.2.1 (#27172)

* chore: bump bytes from 1.1.0 to 1.2.1

Bumps [bytes](https://github.com/tokio-rs/bytes) from 1.1.0 to 1.2.1.
- [Release notes](https://github.com/tokio-rs/bytes/releases)
- [Changelog](https://github.com/tokio-rs/bytes/blob/master/CHANGELOG.md)
- [Commits](tokio-rs/bytes@v1.1.0...v1.2.1)

---
updated-dependencies:
- dependency-name: bytes
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

* [auto-commit] Update all Cargo lock files

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot-buildkite <dependabot-buildkite@noreply.solana.com>

* Share Ancestors API get with contains_key (#27161)

consolidate similar fns

* Rename to `MAX_BLOCK_ACCOUNTS_DATA_SIZE_DELTA` (#27175)

* chore: bump libc from 0.2.129 to 0.2.131 (#27162)

* chore: bump libc from 0.2.129 to 0.2.131

Bumps [libc](https://github.com/rust-lang/libc) from 0.2.129 to 0.2.131.
- [Release notes](https://github.com/rust-lang/libc/releases)
- [Commits](rust-lang/libc@0.2.129...0.2.131)

---
updated-dependencies:
- dependency-name: libc
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

* [auto-commit] Update all Cargo lock files

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot-buildkite <dependabot-buildkite@noreply.solana.com>

* reverts wide fanout in broadcast when the root node is down (#26359)

A change included in
#20480
was that when the root node in turbine broadcast tree is down, the
leader will broadcast the shred to all nodes in the first layer.
The intention was to mitigate the impact of dead nodes on shreds
propagation, because if the root node is down, then the entire cluster
will miss out the shred.
On the other hand, if x% of stake is down, this will cause 200*x% + 1
packets/shreds ratio at the broadcast stage which might contribute to
line-rate saturation and packet drop.
To avoid this bandwidth saturation issue, this commit reverts that logic
and always broadcasts shreds from the leader only to the root node.
As before we rely on erasure codes to recover shreds lost due to staked
nodes being offline.

* add getTokenLargestAccounts rpc method to rust client (#26840)

* add get token largest accounts rpc call to client

* split to include with commitment

* Bump spl-token-2022 (#27181)

* Bump token-2022 to 0.4.3

* Allow cargo to bump stuff to v1.11.5

* VoteProgram.safeWithdraw function to safeguard against accidental vote account closures (#26586)

feat: safe withdraw function

Co-authored-by: aschonfeld <andrew@proofofalpha.io>

* chore: bump futures from 0.3.21 to 0.3.23 (#27182)

* chore: bump futures from 0.3.21 to 0.3.23

Bumps [futures](https://github.com/rust-lang/futures-rs) from 0.3.21 to 0.3.23.
- [Release notes](https://github.com/rust-lang/futures-rs/releases)
- [Changelog](https://github.com/rust-lang/futures-rs/blob/master/CHANGELOG.md)
- [Commits](rust-lang/futures-rs@0.3.21...0.3.23)

---
updated-dependencies:
- dependency-name: futures
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

* [auto-commit] Update all Cargo lock files

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot-buildkite <dependabot-buildkite@noreply.solana.com>

* chore: bump nix from 0.24.2 to 0.25.0 (#27179)

* chore: bump nix from 0.24.2 to 0.25.0

Bumps [nix](https://github.com/nix-rust/nix) from 0.24.2 to 0.25.0.
- [Release notes](https://github.com/nix-rust/nix/releases)
- [Changelog](https://github.com/nix-rust/nix/blob/master/CHANGELOG.md)
- [Commits](nix-rust/nix@v0.24.2...v0.25.0)

---
updated-dependencies:
- dependency-name: nix
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

* [auto-commit] Update all Cargo lock files

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot-buildkite <dependabot-buildkite@noreply.solana.com>

* Parse ConfidentialTransaction instructions (#26825)

Parse ConfidentialTransfer instructions

* snapshots: serialize version file first (#27192)

serialize version file first

* serialize incremental_snapshot_hash (#26839)

* serialize incremental_snapshot_hash

* pr feedback

* derives Error trait for ClusterInfoError and core::result::Error (#27208)

* Add clean_accounts_for_tests() (#27200)

* Rust v1.63.0 (#27148)

* Upgrade to Rust v1.63.0

* Add nightly_clippy_allows

* Resolve some new clippy nightly lints

* Increase QUIC packets completion timeout

Co-authored-by: Michael Vines <mvines@gmail.com>

* docs: updated "transaction fees" page (#26861)

* docs: transaction fees, compute units, compute budget

* docs: added messages definition

* Revert "docs: added messages definition"

This reverts commit 3c56156.

* docs: added messages definition

* Update docs/src/transaction_fees.md

Co-authored-by: Jacob Creech <82475023+jacobcreech@users.noreply.github.com>

* fix: updates from feedback

Co-authored-by: Jacob Creech <82475023+jacobcreech@users.noreply.github.com>

* sdk: Fix args after "--" in build-bpf and test-bpf (#27221)

* Flaky Unit Test test_rpc_subscriptions (#27214)

Increase unit test timeout from 5 seconds to 10 seconds

* chore: only buildkite pipelines use sccache in docker-run.sh (#27204)

chore: only buildkite ci use sccache

* clean feature: `prevent_calling_precompiles_as_programs` (#27100)

* clean feature: prevent_calling_precompiles_as_programs

* fix tests

* fix test

* remove comment

* fix test

* feedback

* Add get_account_with_commitment to BenchTpsClient (#27176)

* Fix a corner-case panic in get_entries_in_data_block() (#27195)

#### Problem
get_entries_in_data_block() panics when there's inconsistency between
slot_meta and data_shred.

However, as we don't lock on reads, reading across multiple column families is
not atomic (especially for older slots) and thus does not guarantee consistency
as the background cleanup service could purge the slot in the middle.  Such
panic was reported in #26980 when the validator serves a high load of RPC calls.

#### Summary of Changes
This PR makes get_entries_in_data_block() panic only when the inconsistency
between slot-meta and data-shred happens on a slot older than lowest_cleanup_slot.

* Verify snapshot slot deltas (#26666)

* store-tool: log lamports for each account (#27168)

log lamports for each account

* add an assert for a debug feature to avoid wasted time (#27210)

* remove redundant call that bumps age to future (#27215)

* Use from_secs api to create duration (#27222)

use from_secs api to create duration

* reorder slot # in debug hash data path (#27217)

* create helper fn for clarity (#27216)

* Verifying snapshot bank must always specify the snapshot slot (#27234)

* Remove `Bank::ensure_no_storage_rewards_pool()` (#26468)

* cli: Add subcommands for address lookup tables (#27123)

* cli: Add subcommand for creating address lookup tables

* cli: Add additional subcommands for address lookup tables

* short commands

* adds hash domain to ping-pong protocol (#27193)

In order to maintain backward compatibility, for now the responding node
will hash the token both with and without domain so that the other node
will accept the response regardless of its upgrade status.
Once the cluster has upgraded to the new code, we will remove the legacy
domain = false case.

* Revert "Rust v1.63.0 (#27148)" (#27245)

This reverts commit a2e7bdf.

* correct double negation (#27240)

* Enable QUIC client by default. Add arg to disable QUIC client. (Forward port #26927) (#27194)

Enable QUIC client by default. Add arg to disable QUIC client.

* Enable QUIC client by default. Add arg to disable QUIC client.
* Deprecate --disable-quic-servers arg
* Add #[ignore] annotation to failing tests

* slots_connected: check if the range is connected (>= ending_slot) (#27152)

* create-snapshot check if snapshot slot exists (#27153)

* Add Bank::clean_accounts_for_tests() (#27209)

* Call `AccountsDb::shrink_all_slots()` directly (#27235)

* add ed25519_program to built-in instruction cost list (#27199)

* add ed25519_program to built-in instruction cost list

* Remove unnecessary and stale comment

* simple refactorings to disk idx (#27238)

* add _inclusive for clarity (#27239)

* eliminate unnecessary ZERO_RAW_LAMPORTS_SENTINEL (#27218)

* make test code more clear (#27260)

* banking stage: actually aggregate tracer packet stats (#27118)

* aggregated_tracer_packet_stats_option was alwasys None

* Actually accumulate tracer packet stats

* Refactor epoch reward 1 (#27253)

* refactor: extract store_stake_accounts fn

* clippy: slice

Co-authored-by: haoran <haoran@mbook>

* recovers merkle shreds from erasure codes (#27136)

The commit
* Identifies Merkle shreds when recovering from erasure codes and
  dispatches specialized code to reconstruct shreds.
* Coding shred headers are added to recovered erasure shards.
* Merkle tree is reconstructed for the erasure batch and added to
  recovered shreds.
* The common signature (for the root of Merkle tree) is attached to all
  recovered shreds.

* Simplify `Bank::clean_accounts()` by removing params (#27254)

* Account files remove (#26910)

* Create a new function cleanup_accounts_paths, a trivial change

* Remove account files asynchronously

* Update and simplify the implementation after the validator test runs.

* Fixes after testing on the dev device

* Discard tokio.  Use thread instead

* Fix comments format

* Fix config type to pass the github test

* Fix failed tests.  Handle the case of non-existing path

* Final cleanup, addressing the review comments
Avoided OsString.
Made the function more generic with "impl AsRef<Path>"

Co-authored-by: Jeff Washington <jeff.washington@solana.com>

* Refactor: Flattens `TransactionContext::instruction_trace` (#27109)

* Flattens TransactionContext::instruction_trace.

* Stop the search at transaction level.

* Renames get_instruction_context_at => get_instruction_context_at_nesting_level.

* Removes TransactionContext::get_instruction_trace().
Adds TransactionContext::get_instruction_trace_length() and TransactionContext::get_instruction_context_at_index().

* Have TransactionContext::instruction_accounts_lamport_sum() accept an iterator instead of a slice.

* Removes instruction_trace from ExecutionRecord.

* make InstructionContext::new() private

* Parallel insertion of dirty store keys during clean (#27058)

parallelize dirty store key insertion

* Refactor epoch reward 2 (#27257)

* refactor: extract store_stake_accounts fn

* refactor: extract store_vote_account fn

* clippy: slice

* clippy: slice

* fix merge error

Co-authored-by: haoran <haoran@mbook>

* Standardize thread names

Tenets:
1. Limit thread names to 15 characters
2. Prefix all Solana-controlled threads with "sol"
3. Use Camel case. It's more character dense than Snake or Kebab case

* cleanup comment on filter_zero_lamport_clean_for_incremental_snapshots (#27273)

* remove inaccurate log (#27255)

* patches metrics for invalid cached vote/stake accounts (#27266)

patches invalid cached vote/stake accounts metrics

Invalid cached vote accounts is overcounting actual mismatches, and
invalid cached stake accounts is undercounting.

* Refactor epoch reward 3 (#27259)

* refactor: extract store_stake_accounts fn

* refactor: extract store_vote_account fn

* refactor: extract reward history update fn

* clippy: slice

* clippy: slice

Co-authored-by: haoran <haoran@mbook>

* fix merges

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: haoran <haoran@mbook>
Co-authored-by: Jeff Biseda <jbiseda@gmail.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: dependabot-buildkite <dependabot-buildkite@noreply.solana.com>
Co-authored-by: Brooks Prumo <brooks@solana.com>
Co-authored-by: behzad nouri <behzadnouri@gmail.com>
Co-authored-by: AJ Taylor <anthonjtaylor@gmail.com>
Co-authored-by: Tyera Eulberg <tyera@solana.com>
Co-authored-by: Andrew Schonfeld <andrew.schonfeld1@gmail.com>
Co-authored-by: aschonfeld <andrew@proofofalpha.io>
Co-authored-by: apfitzge <apfitzge@users.noreply.github.com>
Co-authored-by: Jeff Washington (jwash) <wash678@gmail.com>
Co-authored-by: Brennan Watt <brennan.watt@solana.com>
Co-authored-by: Michael Vines <mvines@gmail.com>
Co-authored-by: Nick Frostbutter <75431177+nickfrosty@users.noreply.github.com>
Co-authored-by: Jacob Creech <82475023+jacobcreech@users.noreply.github.com>
Co-authored-by: Jon Cinque <jon.cinque@gmail.com>
Co-authored-by: Yihau Chen <a122092487@gmail.com>
Co-authored-by: Justin Starry <justin@solana.com>
Co-authored-by: kirill lykov <kirill.lykov@solana.com>
Co-authored-by: Yueh-Hsuan Chiang <93241502+yhchiang-sol@users.noreply.github.com>
Co-authored-by: leonardkulms <42893075+leonardkulms@users.noreply.github.com>
Co-authored-by: Will Hickey <will.hickey@solana.com>
Co-authored-by: Tao Zhu <82401714+taozhu-chicago@users.noreply.github.com>
Co-authored-by: Xiang Zhu <xzhu70@gmail.com>
Co-authored-by: Jeff Washington <jeff.washington@solana.com>
Co-authored-by: Alexander Meißner <AlexanderMeissner@gmx.net>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jeffwashington jeffwashington jeffwashington left review comments

@sakridge sakridge Awaiting requested review from sakridge

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@bw-solana @jeffwashington