MySQL Performance Schema, Open Source India, 2015

Copyright © 2015, Oracle and/or its affiliates. All rights reserved. |
MySQL Performance Schema
Out of the Box in MySQL 5.7
Copyright © 2015, Oracle and/or its affiliates. All rights reserved.
Mayank Prasad
Principal Member Technical Staff
Oracle, MySQL
November 20, 2015

Safe Harbor Statement
The following is intended to outline our general product direction. It is intended for
information purposes only, and may not be incorporated into any contract. It is not a
commitment to deliver any material, code, or functionality, and should not be relied upon
in making purchasing decisions. The development, release, and timing of any features or
functionality described for Oracle’s products remains at the sole discretion of Oracle.
2

Program Agenda
Need and Design
Instruments and instrumentation
Statistics tables
Use cases
What’s new in MySQL 5.7
1
2
3
4
5
3

Program Agenda
Need and Design
Statistics tables
Use cases
1
2
3
4
5
4

Why Performance Schema?
System
Low throughput?
?
5
DBA
Hot table?
Network
High traffic on link?
Storage
Too much Disk spin?App Developer
Slow application?
MySQL Developer
Code contention?
End User
stuck session?

MySQL 5.7 Performance Schema : Design
Block Diagram
MySQL
Server
Instrumentation points
(P_S hooks)
P_S Internal Buffers
P_S
Storage
Engine
Storage
Engine
Interface
Statistics
Report
Fetch
Data
SQL Query
Collect Data
Store
Data
P_S
Tables
6

Program Agenda
Need and Design
Statistics tables
Use cases
1
2
3
4
5
7

Instruments
8
• Name of monitored activity.
• Tree like structure. Separated by ‘/’.
• Left to right : More generic to more specific.
• 1000+ instruments in MySQL 5.7.
• Stored in setup_instruments table.
wait/io/file/myisam/log
statement/sql/select

Program Agenda
Need and Design
Statistics tables
Use cases
1
2
3
4
5
9

Statistics Tables in Performance Schema
10
SETUP
TABLES
Instruments
Actors
Objects
Consumers
TIMERS
EVENTS
TABLES
Transactions
Statements
Stages
Waits
Idle
REPLICATION
SUMMARY
SYSTEM
VARIABLES
STATUS
VARIABLES
LOCK TABLES
Metadata
locks
Table Handles
SUMMARY
TABLES
Events
Memory
File I/O,
Table I/O,
Table locks
Socket
Connection
…
CONNECTION
Attribute
Type
INSTANCE
TABLES
Mutex
RW_locks
File
Sockets
Cond MISC
By_global By_thread By_user/host By_account By_digest

Program Agenda
Need and Design
Statistics tables
Use cases
1
2
3
4
5
11

What does Performance Schema provide …
update setup_instruments set ENABLED='YES', TIMED='YES';
12
Connection 1 (Thread 24)
start transaction;
insert into test.t1 values('11');
commit;
start transaction;
commit;
start transaction;
commit;
select * from test.t1;
Connection 2 (Thread 25)
start transaction;
commit;
start transaction;
commit;
Three
inserts
Two
inserts
Latest
Statement

What does Performance Schema provide … (cont.)
Statements Statistics
* Timer unit is PICOSECOND.
EVENTS_STATEMENTS_CURRENT
THREAD_ID 24 25
EVENT_NAME
statement/sql
/select
statement/sql/
commit
TIMER_WAIT 876585000 15998287000
SQL_TEXT
select * from
test.t1
commit
ROWS_SENT 3 0
NO_INDEX_USED 0 0
SELECT_SCAN 1 0
13
EVENTS_STATEMENTS_SUMMARY_BY_THREAD_BY_EVENT
_NAME
THREAD_ID 24 25
EVENT_NAME
statement/sql
/insert
statement/sql
/insert
COUNT_STAR 3 2
SUM_TIMER_WAIT 35181659000 3477432000
SUM_ROWS_AFFECTED 3 2
SUM_SELECT_SCAN 0 0
SUM_NO_INDEX_USED 0 0

What does Performance Schema provide … (cont.)
Statements Statistics (cont.)
* Timer unit is PICOSECOND.
14
EVENTS_STATEMENTS_SUMMARY_GLOBAL_BY_EVENT_NAME
EVENT_NAME statement/sql/insert statement/sql/commit
COUNT_STAR 5 5
SUM_TIMER_WAIT 38659091000 65812216000
… …
SUM_ROWS_AFFECTED 5 0
… … …

Use case 1
Description
15
• Multiple queries running for long on MySQL Server
• Few long running query (taking lots of time)
• No idea which one
• No idea why
• What to do ? …

Use case 1
Diagnosis
16
– THREAD_ID: 25
– EVENT_ID: 89
– EVENT_NAME: statement/sql/select
– SQL_TEXT : select bla bla bla…;
• Wait ! There’s more!
– SELECT_SCAN : 1
– NO_INDEX_USED: 1
• Aha !!
SELECT * FROM events_statements_history WHERE TIMER_WAIT > ‘X’;

Use case 2
Statements giving errors ( or warnings)
17
SELECT DIGEST_TEXT, SCHEMA_NAME, COUNT_STAR, SUM_ERRORS, SUM_WARNINGS
FROM events_statements_summary_by_digest WHERE SUM_ERRORS > 0
ORDER BY SUM_ERRORS DESC limit 1G;
EVENTS_STATEMENTS_SUMMARY_BY_DIGEST
DIGEST_TEXT CREATE TEMPORARY TABLE IF NOT ... _logs` ( `id` INT8 NOT NULL )!
SCHEMA_NAME mem
COUNT_STAR 1725
SUM_ERRORS 1725
SUM_WARNINGS 0
FIRST_SEEN 2014-05-20 10:42:32
LAST_SEEN 2014-05-21 18:39:22

Use case 3
Description
18
• Multithreaded environment
• My session is stuck
• No idea why
• What to do ? …

Use case 3
• What T1 is waiting for
– Say T1 is waiting for mutex_A (column OBJECT_INSTANCE_BEGIN)
• Lets see who has taken this mutex_A
– Ok, so thread T2 is holding mutex_A (column LOCKED_BY_THREAD_ID)
• Find out what thread t2 is waiting for
• And so on…
SELECT * FROM mutex_instances WHERE OBJECT_INSTANCE_BEGIN = mutex_A;
SELECT * FROM events_waits_current WHERE THREAD_ID = T2;
Diagnosis
19
SELECT * FROM events_waits_current WHERE THREAD_ID = T1;

Replication Summary Tables
Tables for replication statistics
20
Replication status
Connection Information Execute Information
Connection Configuration Connection Status
Execute Configuration Execute Status
Coordinator/SQL
Thread
Worker Thread
replication_connection
_configuration
replication_connection
_status
replication_execute_status
_by_coordinator
replication_execute_status
_by_worker
replication_execute
_status
replication_execute
_configuration

Table Information
replication_connection_configuration (Host, Port, User etc.)
replication_connection_status
(Server UUID, Receiver thread ID, Service
State etc.)
replication_execute_configuration (Desired Delay)
replication_execute_status (Remaining Delay)
replication_execute_status_by_coordinator (Thread Id, Service State, Last Error info.)
replication_execute_status_by_worker (WID, WTID, Service State, Last error Info.)
21

22
SHOW SLAVE STATUS (Limitations)
– No logical division of information.
– Lots of information packed together.
– No cherry picking (difficult for automation).
– Difficult to scale (more new fields).
Why Performance Schema Tables?
– Split logically into different tables.
– SQL Interface. Fetch what is required.
– Easier to extend.

Event Hierarchy
23
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
event_id
nesting_event_id
Transactions Statements Stages Waits

Program Agenda
Need and Design
Statistics tables
Use cases
1
2
3
4
5
24

2525
EnhancementsNew
Instruments
transactionsMemory
usage
Stored
programs
Prepared
statements
Metadata
locks
Connection
type InnoDB
Stages
User
variables
Replication
summary
tables
History per
session
Scalable
memory
allocation
Configurable
digest size
…
MySQL 5.7
Global/Session
variables/status
87 Tables and 1000+ Instruments

SYS Schema (earlier known as P_S Helper)
Procedures Functions Views

Thank You

MySQL Performance Schema, Open Source India, 2015

Recommended

More Related Content

What's hot (20)

Viewers also liked (14)

Similar to MySQL Performance Schema, Open Source India, 2015 (20)

Recently uploaded (20)

MySQL Performance Schema, Open Source India, 2015

Editor's Notes