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

planner: "for update" should not work in a single autocommit statement #11715

Merged
merged 3 commits into from
Aug 13, 2019
Merged

planner: "for update" should not work in a single autocommit statement #11715

merged 3 commits into from
Aug 13, 2019

Conversation

tiancaiamao
Copy link
Contributor

What problem does this PR solve? 

mysql> show create table t;
+-------+------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table                                                                                                                 |
+-------+------------------------------------------------------------------------------------------------------------------------------+
| t     | CREATE TABLE `t` (
  `id` int(11) NOT NULL,
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin |
+-------+------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.01 sec)

mysql> select * from t where id = 9 for update;
ERROR 2013 (HY000): Lost connection to MySQL server during query

What is changed and how it works?

The bug is introduced in this PR #10972

github.com/pingcap/tidb/server.(*clientConn).Run.func1
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:600
runtime.gopanic
	/media/genius/OS/project/go/src/runtime/panic.go:522
github.com/pingcap/tidb/server.(*clientConn).writeResultset.func1
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:1268
runtime.gopanic
	/media/genius/OS/project/go/src/runtime/panic.go:522
runtime.panicmem
	/media/genius/OS/project/go/src/runtime/panic.go:82
runtime.sigpanic
	/media/genius/OS/project/go/src/runtime/signal_unix.go:390
github.com/pingcap/tidb/session.(*TxnState).LockKeys
	<autogenerated>:1
github.com/pingcap/tidb/executor.(*PointGetExecutor).lockKeyIfNeeded
	/media/genius/OS/project/src/github.com/pingcap/tidb/executor/point_get.go:152
github.com/pingcap/tidb/executor.(*PointGetExecutor).Next
	/media/genius/OS/project/src/github.com/pingcap/tidb/executor/point_get.go:132
github.com/pingcap/tidb/executor.Next
	/media/genius/OS/project/src/github.com/pingcap/tidb/executor/executor.go:200
github.com/pingcap/tidb/executor.(*recordSet).Next
	/media/genius/OS/project/src/github.com/pingcap/tidb/executor/adapter.go:113
github.com/pingcap/tidb/server.(*tidbResultSet).Next
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/driver_tidb.go:365
github.com/pingcap/tidb/server.(*clientConn).writeChunks
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:1315
github.com/pingcap/tidb/server.(*clientConn).writeResultset
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:1281
github.com/pingcap/tidb/server.(*clientConn).handleQuery
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:1198
github.com/pingcap/tidb/server.(*clientConn).dispatch
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:897
github.com/pingcap/tidb/server.(*clientConn).Run
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/conn.go:652
github.com/pingcap/tidb/server.(*Server).onConn
	/media/genius/OS/project/src/github.com/pingcap/tidb/server/server.go:438

The root cause of the panic is the mismatch of the life span between statement and executor.
We close the statement (or Txn in the auto-commit case) after Execute returns a RecordSet.
However, RecordSet.Next is still using the Txn, so the panic occurs.

This fix is actually a workaround: change select ... for update statement to normal select ... statement in the auto-commit case.

Check List

Tests

  • Unit test

Related changes

  • Need to cherry-pick to the release branch

@tiancaiamao tiancaiamao added type/bugfix This PR fixes a bug. sig/planner SIG: Planner require-LGT3 Indicates that the PR requires three LGTM. labels Aug 12, 2019
@tiancaiamao
Copy link
Contributor Author

PTAL @coocood @jackysp @lysu

@coocood
Copy link
Member

coocood commented Aug 12, 2019

LGTM

@codecov
Copy link

codecov bot commented Aug 12, 2019 •
edited
Loading

Codecov Report

Merging #11715 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #11715   +/-   ##
===========================================
  Coverage   81.4966%   81.4966%           
===========================================
  Files           432        432           
  Lines         93194      93194           
===========================================
  Hits          75950      75950           
  Misses        11839      11839           
  Partials       5405       5405

@zz-jason zz-jason requested a review from jackysp August 13, 2019 02:55
zz-jason
zz-jason reviewed Aug 13, 2019
View reviewed changes
planner/core/point_get_plan.go
@@ -95,6 +95,9 @@ func (p *PointGetPlan) ExplainInfo() string {
fmt.Fprintf(buffer, ", handle:%d", p.Handle)
}
}
if p.Lock {
fmt.Fprintf(buffer, ", lock")
Copy link
Member

Choose a reason for hiding this comment

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

could you print more information about this lock?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

For example?

jackysp
jackysp previously approved these changes Aug 13, 2019
View reviewed changes
Copy link
Member

@jackysp jackysp left a comment

Choose a reason for hiding this comment

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

LGTM

@jackysp jackysp dismissed their stale review August 13, 2019 06:22

need 3 lgtms

lysu
lysu approved these changes Aug 13, 2019
View reviewed changes
Copy link
Contributor

@lysu lysu left a comment

Choose a reason for hiding this comment

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

LGTM

@coocood coocood added the status/can-merge Indicates a PR has been approved by a committer. label Aug 13, 2019
@sre-bot
Copy link
Contributor

sre-bot commented Aug 13, 2019

/run-all-tests

@sre-bot sre-bot merged commit e08e939 into pingcap:master Aug 13, 2019
@tiancaiamao tiancaiamao deleted the point-get-for-update branch August 14, 2019 02:39
tiancaiamao added a commit to tiancaiamao/tidb that referenced this pull request Aug 14, 2019
@tiancaiamao
planner: "for update" should not work in a single autocommit statement (
78a47dd 
pingcap#11715)
@tiancaiamao tiancaiamao mentioned this pull request Aug 14, 2019
Closed
@sre-bot sre-bot mentioned this pull request Aug 14, 2019
Merged
@sre-bot
Copy link
Contributor

sre-bot commented Aug 14, 2019

cherry pick to release-3.0 in PR #11736

@tiancaiamao tiancaiamao mentioned this pull request Aug 19, 2019
Merged
@sre-bot sre-bot mentioned this pull request Aug 19, 2019
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@zz-jason zz-jason zz-jason left review comments

@lysu lysu lysu approved these changes

@jackysp jackysp jackysp left review comments

Assignees
No one assigned
Labels
require-LGT3 Indicates that the PR requires three LGTM. sig/planner SIG: Planner status/can-merge Indicates a PR has been approved by a committer. type/bugfix This PR fixes a bug.
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
6 participants
@tiancaiamao @coocood @sre-bot @lysu @jackysp @zz-jason