MySQL InnoDB Cluster and Group Replication - OSI 2017 Bangalore

MySQL InnoDB Cluster and Group Replication

Madhusudhan G Joshi([email protected])Software Development Manager

Sujatha Sivakumar([email protected])Principal Member Technical Staf

Copyright © 2017, Oracle and/or its affiliates. All rights reserved. | OSI, Bangalore, 13 October 2017 2

Safe Harbour 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.


Program Agenda

Introduction to MySQL Innodb Cluster

How to get it working

Introduction to Group Replication

Look and Feel of GR!

Monitoring

1

2

3

4

5


Introduction to MySQL Innodb Cluster1


InnoDB ClusterIntegrated Solution● Group Replication for high availability.● Asynchronous Replication for Read Scale-out.● One-stop shell to deploy and manage the cluster.● Seamlessly and automatically route the workload to the proper database server in the cluster.● Hide failures from the application.

M

M M

MySQL Connector

Application

MySQL Router

MySQL Connector

Application

MySQL Router

MQL Shell

HA

S1 S2 S3 S4 S…


How to get it working2


MySQL InnoDB Cluster

• Start MySQL Shell• Deploy Instance (s)• dba.deploySandboxInstance(....)



• Start MySQL Shell• Deploy Instance (s)

• Initiate/Create ClusterInitiate/Create Cluster

shell.connect("root@localhost:1310");shell.connect("root@localhost:1310");

cluster = dba.createCluster('OSIcluster');cluster = dba.createCluster('OSIcluster');



few more moments ….

Is the Cluster on and available ?



• Start MySQL Shell• Deploy Instance (s)• Create Cluster

Add few more instancesAdd few more instances

myOSI.addInstance("root@localhost:XXXX");myOSI.addInstance("root@localhost:XXXX");


MySQL InnoDB ClusterHow is the InnoDB Cluster now ?


MySQL InnoDB Cluster• Start MySQL Shell• Deploy Instance (s)• Create Cluster• Add more instances

Setup and start RouterSetup and start Routerbin/mysqlrouter --bootstrap localhost:bin/mysqlrouter --bootstrap localhost:

1310 --name OSIRouter1310 --name OSIRouter

Notice the R/W, R ports for MySQL

InnoDB Cluster



MySQL Router( layered on top of the MySQL Instances )

‘R’ InstancePort :1311

‘R/W’ InstancePort :1310


Router is set and HA’d


MySQL InnoDB Cluster…on test

• Start MySQL Shell• Deploy Instance (s)• Create Cluster• Add more instances• Setup and Start Router

Client connects to R/W Instance Client connects to R/W Instance

and and

Kill R/W InstanceKill R/W Instance

Where do new client requests go to ???


MySQL InnoDB Cluster…on test

• Start MySQL Shell• Deploy Instance (s)• Create Cluster• Add more instances• Setup and Start Router• Client Connects to R/W Instance• Kill R/W Instance

Client Requests handled by new R/W InstanceClient Requests handled by new R/W Instance

How does the InnoDB Cluster look like ….now ???



MySQL Router( layered on top of the MySQL Instances )

‘R/W’ InstancePort :1311


Copyright © 2017, Oracle and/or its affiliates. All rights reserved. | Percona Live, Dublin, 27 September 2017 17

Introduction to Group Replication3


MySQL Group ReplicationTraditional database● Many clients● One server

server

clients


MySQL Group Replication

serverA

serverB

serverC

clientsGroup Replication● Many clients● Many servers● Read scalability● Write everywhere*● Elastic,

Highly available● Agreement on:

● Members● Message delivery● State updates



serverA

serverB

W WW

serverC

R RR R RR






serverA

serverB

W RW

serverC

R W RW

clients

WR

Group Replication● Many clients● Many servers● Read scalability● Write everywhere*● Elastic,



Limitations apply!https://dev.mysql.com/doc/refman/8.0/en/group-replication-limitations.html



serverA

serverB

W RW

serverC

R W RW

clients

serverD

WR






serverA

serverB

W RW

serverC

WR R W RW

clients

serverD

serverE






serverA

serverB

W RW

serverC

R WR W RW

clients

serverD

serverE






serverA

serverB

W RW

serverC

R WR W RW

clients

serverD

serverE






serverA

serverB

W RW

serverC

R WR W RW

clients

serverE






serverA

serverB

W RW

serverC

R WR W RW

clients

serverE






serverA

serverB

W RW

serverC

R WR W RW





8.0 Operations: Set Next Primary

B

A

C

W WW

SET group_replication_member_weight = [0-100]

Dw=25 w=20

R RR R RR R RR

w=10w=35


W WW


B

A

Cw=35

Dw=10 w=25 w=20

R RR R RR R RR




B

C

Dw=10 w=25 w=20

? ? ?

W WW R RR R RR R RR





B

C

Dw=10 w=25 w=20

? ? ?

W WW R RR R RR R RR

25 is the biggest→ C is new primary



B

C

Dw=10 w=25 w=20

W WW R RR R RR R RR



8.0 Operations: No Writes After Leaving Group

B

A

C

D

W WW



B

A

C

D

Locally on A:

STOP GROUP REPLICATION

W WW

1. Disallow new writes2. Wait for ongoing writes3. Leave group



B

A

C

D

Locally on A:

STOP GROUP REPLICATION

W WW



W WW


B

A

C

D

Locally on A:STOP GROUP REPLICATION




serverA

serverB

W RW

serverC

R WR W RW

clients

We are 3:A, B, C!

We are 3:A, B, C!

We are 3:A, B, C!






serverA

serverB

W RW

serverC

R WR W RW






serverA

serverB

W RW R WR W RW

clients

You and me:A and B!

You and me:A and B!






serverA

serverB

W RW R WR W RW




B B A A

All receive insame order



serverA

serverB

W RW R WR W RW

clients




Tablerow 1row 2row 3

Queue

insert row 4

Tablerow 1row 2

Queue

insert row 3insert row 4



serverA

serverB

W RW WR

clients

Group Replication● Non conflicting

transactions are

certified positive.● For Conflicting

transactions firstcommitter wins

UPDATE t1 SET a=3

WHERE a=1

UPDATE t1SET a=4

Where a=2

OK OK



serverA

serverB

W RW WRclients

Group Replication● Non conflicting

transactions are

certified positive.● For Conflicting

transactions firstcommitter wins

UPDATE t1 SET a=3

WHERE a=1 UPDATE t1SET a=4

Where a=1OK


Look and Feel of GR!4


MySQL Group ReplicationGroup Replication● MySQL plugin● Uses InnoDB● Uses row format● Monitor using

performance schema● Full GTID support● Auto-increment

handling● Version access control

GR in 8 is gr8


MySQL Group ReplicationGroup Replication● MySQL plugin● Uses InnoDB● Monitor using

performance schema●

● Auto-increment

handling● Version access control

serverA

serverB

W RW WR

clients

INSERT x;Will have

GTID:UUID:1

INSERT y;Will have

GTID:UUID:2

Full GTID support

All group members share the same UUID, the group name.



performance schema● Full GTID support●

● Version access control

serverA

serverB

W RW WR

INSERT x;X:1

INSERT y;Y:2

INSERT z;Y:9

Auto-increment handling

●By default, the ofset is provided by server_id and increment is 7.●Users can change the increment size usinggroup_replication_auto_increment_increment

option.




handling

Version access control

MySQL

5.7.13

MySQL

5.7.14

W RW

MySQL

5.7.15

R WR W RW

Member with different patch version is allowed to join the gorup.

MySQL

5.7.16

OK




handling


MySQL

8.0

MySQL

8.0

W RW

MySQL

8.0

R WR W RW

Member with lower major version than the major version in the group is not allowed to join group.

MySQL

5.7.16

No




handling


MySQL

5.7.13

MySQL

5.7.14

W RW

MySQL

5.7.15

R WR W RW

Member with higher major version is allowed to join but is not allowed to do writes. (Read only)

MySQL

8.0

OK


MySQL Async Replication


MySQL Async Replication

Example: Async replication between geo-distributed groups


Replication Architecture: Applier Pipeline

Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network



Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network


8.0 Performance: Applier Pipeline Parallelize reads and writes to relay log

Performance for free

Replica

RelayLog

WorkerQueues

WorkerThreads

Coordinator Thread

ReceiverThread

network

trx1

trx1


Monitoring5


8.0 Monitoring: Group Members Who are in this group?

– server_uuid, host, port, MySQL version

Who is the primary? Online status of each node

mysql> SELECT … FROM performance_schema.replication_group_membersMEMBER_ID MEMBER_HOST MEMBER_PORT MEMBER_STATE MEMBER_ROLE MEMBER_VERSION<uuid-A> murkla 3306 ONLINE PRIMARY 5.7.20<uuid-B> kremla 3306 ONLINE SECONDARY 5.7.21<uuid-C> kantarell 3306 UNREACHABLE SECONDARY 8.0.3

New in 8.0

B5.7.21

A5.7.20

C8.0.3


8.0 Monitoring: Group Member Statistics Statistics used to diagnose group members

– Conflict ratio of the group– Conflict ratio of each member (multi-primary)– Is any member falling behind on certification or applying?

Table performance_schema.replication_group_member_stats:counts of transactions, by member:

– Positively/negatively certified transactions of any origin– Positively/negatively certified transactions from the node– Transactions waiting to be certified– Certified transactions waiting to be applied


“I can’t w8 to replic8 with version 8” Evalu8

http://www.mysql.com/downloads/

Educ8Innodb cluster:

https://dev.mysql.com/doc/refman/8.0/en/mysql-innodb-cluster-userguide.html

group replication:

http://dev.mysql.com/doc/refman/8.0/en/group-replication.html

Contempl8http://mysqlhighavailability.com

Cooper8https://bugs.mysql.com/


Thank You!& cu l8r, m8s

Copyright © 2012, Oracle and/or its affiliates. All rights reserved. | 81