Skip to content

[FLASH-572] Use jemalloc instead of tcmalloc by default #424

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
Merged
JaySon-Huang merged 11 commits into from
Feb 12, 2020
Merged

[FLASH-572] Use jemalloc instead of tcmalloc by default #424

JaySon-Huang merged 11 commits into from
Feb 12, 2020

Conversation

@JaySon-Huang
Copy link
Contributor

@JaySon-Huang JaySon-Huang commented Feb 10, 2020 •
edited
Loading

  • Possible tcmalloc have bugs and make tiflash stuck to refuse connections. We have happened to to this twice
  • Clickhouse use jemalloc instead of tcmalloc, since jemalloc have better performance without tuning special params. And tcmalloc is no longer supported. Using jemalloc instead of tcmalloc. ClickHouse/ClickHouse#2773

This PR

  • Port contrib/jemalloc,contrib/jemalloc-cmake from clickhouse and use jemalloc under Linux by default
  • Use tcmalloc under MacOS by default (jemalloc will cause compile error under MacOS)
  • Fix bugs: use after free
    • TMTContext should be free before background_pool.
      Some background handlers need to be removed in TMTContext's BackgroundService and Storages
      Context.cpp:L148
    • Use of invalid iterator in AlterCommand
      AlterCommands.cpp:L186

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 10, 2020

/run-integration-tests

@JaySon-Huang
Use jemalloc in Linux and tcmalloc in MaxOS
ccc6639 
Signed-off-by: JaySon-Huang <jayson.hjs@gmail.com>
@JaySon-Huang
Fix bug: TMTContext should be free before background_pool
d103f11 
Signed-off-by: JaySon-Huang <jayson.hjs@gmail.com>
@JaySon-Huang
Merge branch 'master' into FLASH-572
989e3f9
@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 10, 2020

/run-integration-tests

2 similar comments
@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 10, 2020

/run-integration-tests

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

ilovesoup
ilovesoup approved these changes Feb 11, 2020
View reviewed changes
Copy link
Contributor

@ilovesoup ilovesoup left a comment

Choose a reason for hiding this comment

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

LGTM

dbms/src/Storages/StorageDeltaMerge.cpp
shutdown();
#if USE_TCMALLOC
// Reclaim memory.
MallocExtension::instance()->ReleaseFreeMemory();
Copy link
Contributor

@ilovesoup ilovesoup Feb 11, 2020

Choose a reason for hiding this comment

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

Why jemalloc does not need this?

Copy link
Contributor Author

@JaySon-Huang JaySon-Huang Feb 11, 2020 •
edited
Loading

Choose a reason for hiding this comment

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

For tcmalloc, clickhouse have to tune "tcmalloc.aggressive_memory_decommit" to be false to get better performance.
This piece of code is just for tcmalloc, after drop table, remind tcmalloc to reclaim free memory, or it acts like to be leak. See ClickHouse/ClickHouse#1742 for details.

But with jemalloc, no tuning is required:

When using jemalloc, there are no such issues even ** without any tuning. **
Some queries will run faster up to 20%.
Memory usage (RSS) is about 10% lower and is more stable.

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

3 similar comments
@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

@JaySon-Huang
Copy link
Contributor Author

JaySon-Huang commented Feb 11, 2020

/run-integration-tests

@JaySon-Huang JaySon-Huang merged commit af17e0d into pingcap:master Feb 12, 2020
@JaySon-Huang JaySon-Huang deleted the FLASH-572 branch February 12, 2020 03:39
@yiguolei yiguolei mentioned this pull request Sep 23, 2022
Merged
13 tasks
yongman added a commit to yongman/tiflash that referenced this pull request Jun 18, 2025
@yongman
proxy: bump to fix vector index support mvcc delete (pingcap#424)
ca57d8f 
Signed-off-by: yongman <yming0221@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@hanfei1991 hanfei1991 Awaiting requested review from hanfei1991

1 more reviewer

@ilovesoup ilovesoup ilovesoup approved these changes

Reviewers whose approvals may not affect merge requirements

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

@JaySon-Huang @ilovesoup