feat(streaming): use tokio channels and better coop scheduling #3374

skyzh · 2022-06-21T06:32:09Z

Signed-off-by: Alex Chi iskyzh@gmail.com

What's changed and what's your intention?

This PR enables unstable feature of tokio, so that we can use budget APIs.
Replace all channels to tokio channels so that they can opt in schedule budgets.
SSTableIterator will account for one budget when seek.

metrics looks good:

Checklist

I have written necessary docs and comments
I have added necessary unit tests and integration tests
All checks passed in ./risedev check (or alias, ./risedev c)

Refer to a related PR or issue link (optional)

skyzh · 2022-06-21T06:33:16Z

The bench command can now yield normally.

codecov · 2022-06-21T06:54:57Z

Codecov Report

Merging #3374 (270d72e) into main (4b33d95) will decrease coverage by 0.00%.
The diff coverage is 79.16%.

@@            Coverage Diff             @@
##             main    #3374      +/-   ##
==========================================
- Coverage   73.63%   73.63%   -0.01%     
==========================================
  Files         760      760              
  Lines      104229   104231       +2     
==========================================
  Hits        76754    76754              
- Misses      27475    27477       +2

Flag	Coverage Δ
rust	`73.63% <79.16%> (-0.01%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/compute/src/rpc/service/exchange_service.rs	`0.00% <0.00%> (ø)`
src/ctl/src/cmd_impl/bench.rs	`0.00% <ø> (ø)`
src/risedevtool/src/bin/risedev-compose.rs	`0.43% <0.00%> (-0.01%)`	⬇️
src/storage/src/hummock/sstable_store.rs	`83.03% <ø> (-0.71%)`	⬇️
src/stream/src/task/mod.rs	`43.75% <ø> (ø)`
src/stream/src/task/stream_manager.rs	`0.00% <0.00%> (ø)`
src/common/src/cache.rs	`96.73% <100.00%> (ø)`
...ge/src/hummock/sstable/forward_sstable_iterator.rs	`95.12% <100.00%> (+0.51%)`	⬆️
src/stream/src/executor/dispatch.rs	`76.94% <100.00%> (ø)`
src/stream/src/executor/integration_tests.rs	`98.40% <100.00%> (-0.54%)`	⬇️
... and 10 more

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

Signed-off-by: Alex Chi <iskyzh@gmail.com>

skyzh · 2022-06-21T08:12:14Z

Also benchmark shows that the state table layer can scan 50000 rows/s per core, which is about 500000 KV/s.

Which looks (maybe) reasonable?

jon-chuang · 2022-06-21T09:38:45Z

src/storage/src/hummock/sstable/forward_sstable_iterator.rs

+        // When all data are in block cache, it is highly possible that this iterator will stay on a
+        // worker thread for a full time. Therefore, we use tokio's unstable API consume_budget to
+        // do cooperative scheduling.
+        tokio::task::consume_budget().await;


is there a way we can achieve something similar without tokio unstable? Maybe we could simply yield when idx % YIELD_ITERATIONS == 0?

Yes, but it won't share the budget with tokio channels / semaphores. I think it would be better if we have a single budget for all related events :)

jon-chuang · 2022-06-21T09:43:44Z

src/stream/src/executor/merge.rs

@@ -136,7 +137,7 @@ pub async fn cooperative_scheduling<T>(stream: impl Stream<Item = T>) {
    #[for_await]
    for item in stream {
        yield item;
-        tokio::task::yield_now().await;
+        // tokio::task::consume_budget().await;


similar comment here

jon-chuang · 2022-06-21T09:50:53Z

Quite curious if this has been benchmarked on the hash join?

skyzh · 2022-06-21T09:56:59Z

Yes, still no significant improvement.

Signed-off-by: Alex Chi <iskyzh@gmail.com>

…into skyzh/use-tokio-channel

Signed-off-by: Alex Chi <iskyzh@gmail.com>

commit aac69d2 Author: William Wen <44139337+wenym1@users.noreply.github.com> Date: Wed Jun 22 17:23:31 2022 +0800 fix(storage): fix ignoring delete record when getting from sst (#3405) Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit ebfbd0c Author: Alex Chi <iskyzh@gmail.com> Date: Wed Jun 22 17:01:53 2022 +0800 chore(storage): move read lock acquire out of lock (#3400) Signed-off-by: Alex Chi <iskyzh@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit c5b4fe2 Author: lmatz <lmatz823@gmail.com> Date: Wed Jun 22 01:49:23 2022 -0700 fix(expr): trim (#3402) commit 28cf5ff Author: Wallace <bupt2013211450@gmail.com> Date: Wed Jun 22 16:29:03 2022 +0800 feat(compaction): compress data with dynamic level (#3388) * do not compress data in high level Signed-off-by: Little-Wallace <bupt2013211450@gmail.com> commit 7602285 Author: Lee Zong Yu <65748142+marvenlee2486@users.noreply.github.com> Date: Wed Jun 22 15:58:31 2022 +0800 fix(grafana): Edit update.sh (#3399) * Edit update.sh * Edit format error * Add payload to gitignore * Revert risedev.yml modification * Fix wrong gitignore * Fix wrong gitignore Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit 4e66ca3 Author: congyi <58715567+wcy-fdu@users.noreply.github.com> Date: Wed Jun 22 13:16:48 2022 +0800 feat(docs): add docs for the relational table layer (#3313) * add doc for relational layer commit 7c44a15 Author: Bugen Zhao <i@bugenzhao.com> Date: Wed Jun 22 12:51:59 2022 +0800 fix: build failure of cell based table with release profile (#3395) commit 93c62a3 Author: Li0k <yuli@singularity-data.com> Date: Wed Jun 22 11:50:18 2022 +0800 chore(frontend): fix handle_with_properties ctx of explain (#3394) commit 233ee5d Author: Li0k <yuli@singularity-data.com> Date: Wed Jun 22 11:32:05 2022 +0800 feat(frontend): catalog add properties for ttl (#3382) * feat(frontend): catlog add properties for materialized_view and materialized_source to support ttl * fix(frontend): fix test_runner * chore(frontend): unify handle_with_properties logic for handler * chore(frontend): explain introduce handle_with_properties to replace WithProperties * chore(sqlparser): remove unused logic of WithProperties commit 78861e3 Author: Bugen Zhao <i@bugenzhao.com> Date: Wed Jun 22 02:19:37 2022 +0800 feat(storage): store all column descs in cell based table (#3344) * add pk indices Signed-off-by: Bugen Zhao <i@bugenzhao.com> * extract mapping Signed-off-by: Bugen Zhao <i@bugenzhao.com> * distinguish parital table Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refactor proto Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix clippy Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix output schema Signed-off-by: Bugen Zhao <i@bugenzhao.com> * store column ids in serializer Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refine docs Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix plan test Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refine docs Signed-off-by: Bugen Zhao <i@bugenzhao.com> * write check Signed-off-by: Bugen Zhao <i@bugenzhao.com> * Apply suggestions from code review Co-authored-by: Yuanxin Cao <60498509+xx01cyx@users.noreply.github.com> * trigger ci Signed-off-by: Bugen Zhao <i@bugenzhao.com> Co-authored-by: Yuanxin Cao <60498509+xx01cyx@users.noreply.github.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit bc74ff4 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 22:59:22 2022 +0800 feat(streaming): use tokio channels and better coop scheduling (#3374) * feat(streaming): use tokio channels and better coop scheduling Signed-off-by: Alex Chi <iskyzh@gmail.com> * more info when deploy Signed-off-by: Alex Chi <iskyzh@gmail.com> * remove manual coop scheduling Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> commit 4b33d95 Author: Huangjw <1223644280@qq.com> Date: Tue Jun 21 18:32:36 2022 +0800 chore(ci): update ci image vesion and remove changelog.json (#3384) Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit ed1de15 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 18:19:02 2022 +0800 feat(risedev): fix meta args and support healthcheck in docker (#3378) * feat(risedev): support healthcheck in docker Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * pin version Signed-off-by: Alex Chi <iskyzh@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit 5a7e115 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 18:05:56 2022 +0800 chore(ci): manual CLA check (#3386) Signed-off-by: Alex Chi <iskyzh@gmail.com> commit 2688491 Author: ZENOTME <43447882+ZENOTME@users.noreply.github.com> Date: Tue Jun 21 18:03:02 2022 +0800 feat(pgwire): support row limit in extended query mode (#3354) * * add row_end flag in PgResponse * add values interface in PgResponse * * add result_cache * add execute interface * fix complile error of row_end * * split process_query_msg into process_query_msg_simple and process_query_response execute flow: simple mode: extended query mode: process_query_msg_simple portal.execute() | (response) (response) | process_query_response --------------------- | process_query_with_results * fix clippy * add process at Execute * add PortalSuspended msg * add row_limit process at process_query_with_results * fix small problem and add some comment Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit dc8631e Author: Tao Wu <wutao@singularity-data.com> Date: Tue Jun 21 17:35:36 2022 +0800 test: sqlsmith supports generating unary func (#3370) commit cb0d7c1 Author: Bowen <36908971+BowenXiao1999@users.noreply.github.com> Date: Tue Jun 21 17:16:10 2022 +0800 refactor: deparallel hash agg apply batch + remove Arc Mutex (#3377) * refactor: deparallel hash agg apply batch * remove Arc Mutex Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

…3363) * implement cascade && restrict by add relation map * fmt * implement cascade && restrict by add relation map * fmt * correct pre unit test error * correct misc test error * fix * update proto && fix * change grantor to granted by * implement recursive privilege relations && fix * Squashed commit of the following: commit aac69d2 Author: William Wen <44139337+wenym1@users.noreply.github.com> Date: Wed Jun 22 17:23:31 2022 +0800 fix(storage): fix ignoring delete record when getting from sst (#3405) Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit ebfbd0c Author: Alex Chi <iskyzh@gmail.com> Date: Wed Jun 22 17:01:53 2022 +0800 chore(storage): move read lock acquire out of lock (#3400) Signed-off-by: Alex Chi <iskyzh@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit c5b4fe2 Author: lmatz <lmatz823@gmail.com> Date: Wed Jun 22 01:49:23 2022 -0700 fix(expr): trim (#3402) commit 28cf5ff Author: Wallace <bupt2013211450@gmail.com> Date: Wed Jun 22 16:29:03 2022 +0800 feat(compaction): compress data with dynamic level (#3388) * do not compress data in high level Signed-off-by: Little-Wallace <bupt2013211450@gmail.com> commit 7602285 Author: Lee Zong Yu <65748142+marvenlee2486@users.noreply.github.com> Date: Wed Jun 22 15:58:31 2022 +0800 fix(grafana): Edit update.sh (#3399) * Edit update.sh * Edit format error * Add payload to gitignore * Revert risedev.yml modification * Fix wrong gitignore * Fix wrong gitignore Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit 4e66ca3 Author: congyi <58715567+wcy-fdu@users.noreply.github.com> Date: Wed Jun 22 13:16:48 2022 +0800 feat(docs): add docs for the relational table layer (#3313) * add doc for relational layer commit 7c44a15 Author: Bugen Zhao <i@bugenzhao.com> Date: Wed Jun 22 12:51:59 2022 +0800 fix: build failure of cell based table with release profile (#3395) commit 93c62a3 Author: Li0k <yuli@singularity-data.com> Date: Wed Jun 22 11:50:18 2022 +0800 chore(frontend): fix handle_with_properties ctx of explain (#3394) commit 233ee5d Author: Li0k <yuli@singularity-data.com> Date: Wed Jun 22 11:32:05 2022 +0800 feat(frontend): catalog add properties for ttl (#3382) * feat(frontend): catlog add properties for materialized_view and materialized_source to support ttl * fix(frontend): fix test_runner * chore(frontend): unify handle_with_properties logic for handler * chore(frontend): explain introduce handle_with_properties to replace WithProperties * chore(sqlparser): remove unused logic of WithProperties commit 78861e3 Author: Bugen Zhao <i@bugenzhao.com> Date: Wed Jun 22 02:19:37 2022 +0800 feat(storage): store all column descs in cell based table (#3344) * add pk indices Signed-off-by: Bugen Zhao <i@bugenzhao.com> * extract mapping Signed-off-by: Bugen Zhao <i@bugenzhao.com> * distinguish parital table Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refactor proto Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix clippy Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix output schema Signed-off-by: Bugen Zhao <i@bugenzhao.com> * store column ids in serializer Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refine docs Signed-off-by: Bugen Zhao <i@bugenzhao.com> * fix plan test Signed-off-by: Bugen Zhao <i@bugenzhao.com> * refine docs Signed-off-by: Bugen Zhao <i@bugenzhao.com> * write check Signed-off-by: Bugen Zhao <i@bugenzhao.com> * Apply suggestions from code review Co-authored-by: Yuanxin Cao <60498509+xx01cyx@users.noreply.github.com> * trigger ci Signed-off-by: Bugen Zhao <i@bugenzhao.com> Co-authored-by: Yuanxin Cao <60498509+xx01cyx@users.noreply.github.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit bc74ff4 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 22:59:22 2022 +0800 feat(streaming): use tokio channels and better coop scheduling (#3374) * feat(streaming): use tokio channels and better coop scheduling Signed-off-by: Alex Chi <iskyzh@gmail.com> * more info when deploy Signed-off-by: Alex Chi <iskyzh@gmail.com> * remove manual coop scheduling Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> commit 4b33d95 Author: Huangjw <1223644280@qq.com> Date: Tue Jun 21 18:32:36 2022 +0800 chore(ci): update ci image vesion and remove changelog.json (#3384) Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit ed1de15 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 18:19:02 2022 +0800 feat(risedev): fix meta args and support healthcheck in docker (#3378) * feat(risedev): support healthcheck in docker Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * fix Signed-off-by: Alex Chi <iskyzh@gmail.com> * pin version Signed-off-by: Alex Chi <iskyzh@gmail.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit 5a7e115 Author: Alex Chi <iskyzh@gmail.com> Date: Tue Jun 21 18:05:56 2022 +0800 chore(ci): manual CLA check (#3386) Signed-off-by: Alex Chi <iskyzh@gmail.com> commit 2688491 Author: ZENOTME <43447882+ZENOTME@users.noreply.github.com> Date: Tue Jun 21 18:03:02 2022 +0800 feat(pgwire): support row limit in extended query mode (#3354) * * add row_end flag in PgResponse * add values interface in PgResponse * * add result_cache * add execute interface * fix complile error of row_end * * split process_query_msg into process_query_msg_simple and process_query_response execute flow: simple mode: extended query mode: process_query_msg_simple portal.execute() | (response) (response) | process_query_response --------------------- | process_query_with_results * fix clippy * add process at Execute * add PortalSuspended msg * add row_limit process at process_query_with_results * fix small problem and add some comment Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> commit dc8631e Author: Tao Wu <wutao@singularity-data.com> Date: Tue Jun 21 17:35:36 2022 +0800 test: sqlsmith supports generating unary func (#3370) commit cb0d7c1 Author: Bowen <36908971+BowenXiao1999@users.noreply.github.com> Date: Tue Jun 21 17:16:10 2022 +0800 refactor: deparallel hash agg apply batch + remove Arc Mutex (#3377) * refactor: deparallel hash agg apply batch * remove Arc Mutex Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> * add unit test * fmt * correct pre test * some fix Co-authored-by: August <pin@singularity-data.com>

github-actions bot added the type/feature label Jun 21, 2022

skyzh force-pushed the skyzh/use-tokio-channel branch from 9c2a364 to 23503eb Compare June 21, 2022 06:42

feat(streaming): use tokio channels and better coop scheduling

a31f5c8

Signed-off-by: Alex Chi <iskyzh@gmail.com>

skyzh force-pushed the skyzh/use-tokio-channel branch from 23503eb to a31f5c8 Compare June 21, 2022 07:34

more info when deploy

8cf9975

Signed-off-by: Alex Chi <iskyzh@gmail.com>

skyzh marked this pull request as ready for review June 21, 2022 08:08

skyzh requested review from Little-Wallace and TennyZhuang June 21, 2022 08:10

Merge branch 'main' into skyzh/use-tokio-channel

e6c3fff

jon-chuang reviewed Jun 21, 2022

View reviewed changes

skyzh requested a review from BugenZhao June 21, 2022 10:53

remove manual coop scheduling

712193a

Signed-off-by: Alex Chi <iskyzh@gmail.com>

BugenZhao approved these changes Jun 21, 2022

View reviewed changes

skyzh added 2 commits June 21, 2022 22:45

Merge branch 'main' of https://github.com/singularity-data/risingwave …

a457258

…into skyzh/use-tokio-channel

fix

270d72e

Signed-off-by: Alex Chi <iskyzh@gmail.com>

skyzh added the mergify/can-merge Indicates that the PR can be added to the merge queue label Jun 21, 2022

mergify bot merged commit bc74ff4 into main Jun 21, 2022

mergify bot deleted the skyzh/use-tokio-channel branch June 21, 2022 14:59

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(streaming): use tokio channels and better coop scheduling #3374

feat(streaming): use tokio channels and better coop scheduling #3374

skyzh commented Jun 21, 2022 •

edited

skyzh commented Jun 21, 2022

codecov bot commented Jun 21, 2022 •

edited

skyzh commented Jun 21, 2022

jon-chuang Jun 21, 2022 •

edited

skyzh Jun 21, 2022

jon-chuang Jun 21, 2022

jon-chuang commented Jun 21, 2022

skyzh commented Jun 21, 2022

feat(streaming): use tokio channels and better coop scheduling #3374

feat(streaming): use tokio channels and better coop scheduling #3374

Conversation

skyzh commented Jun 21, 2022 • edited

What's changed and what's your intention?

Checklist

Refer to a related PR or issue link (optional)

skyzh commented Jun 21, 2022

codecov bot commented Jun 21, 2022 • edited

Codecov Report

skyzh commented Jun 21, 2022

jon-chuang Jun 21, 2022 • edited

Choose a reason for hiding this comment

skyzh Jun 21, 2022

Choose a reason for hiding this comment

jon-chuang Jun 21, 2022

Choose a reason for hiding this comment

jon-chuang commented Jun 21, 2022

skyzh commented Jun 21, 2022

skyzh commented Jun 21, 2022 •

edited

codecov bot commented Jun 21, 2022 •

edited

jon-chuang Jun 21, 2022 •

edited