From single MySQL instance to High Availability: the journey to MySQL InnoDB Cluster

Frédéric Descamps
Community Manager
MySQL
February 2021
From Single MySQL
Instance to HA
The Journey To MySQL InnoDB
Cluster

Who am I ?
about.me/lefred
Copyright @ 2021 Oracle and/or its affiliates.
2

Frédéric Descamps
@lefred
MySQL Evangelist
Managing MySQL since 3.20
devops believer
living in Belgium
h ps://lefred.be
hairdressers are closed in .be
3

Evolution to HA
MySQL 8.0
4

5

This is where it all begins...
Some tips:
use exclusively InnoDB
keep the durability defaults
Single MySQL Instance

6

Next level
The database becomes more important, loosing it might be an issue...
7

Next level
RTO → hours
7

Next level
RTO → hours

RPO → 1 day
7

Next level
RTO → hours

RPO → 1 day
RTO: Recovery Time Objective (how long to recover)
RPO: Recovery Point Objective (how much data can be lost)
7

Physical Backups
Logical Backups
For logical backups, please use MySQL Shell
Dump & Load Utility !
Backups
8

Next level
RPO of 1 day ? Really ?? We want to reduce it to minutes at least !
9

Next level
RTO → hours
9

Next level
RTO → hours

RPO → minutes
9

This is the default in MySQL 8.0:
Durable Binlogs

10

Next level
My data is very important and I've a heavy workload... I would like to loose less than a
second !
11

Next level
second !
RTO → hours
11

Next level
second !
RTO → hours

RPO → less than a second
11

Enable GTID (optional but so convenient)
O -load binlogs (in real time)
Point-In-Time Recovery (PTR)
12

with restart
Enable GTID
13

with restart without restart
Enable GTID
13

O -load Binlogs

14

O -load Binlogs

On another machine:
mysqlbinlog --raw --read-from-remote-server --stop-never --host 10.0.0.2
--port 3306 -u getbinlog -ppassword
--ssl-mode='REQUIRED' binlog.xxxxxx
14

O -load Binlogs

On another machine:
mysqlbinlog --raw --read-from-remote-server --stop-never --host 10.0.0.2
--port 3306 -u getbinlog -ppassword
--ssl-mode='REQUIRED' binlog.xxxxxx
See h ps://lefred.be/content/howto-make-mysql-point-in-time-recovery-faster/
14

Next level
My service is important, I would like to be up again in some minutes
15

Next level
RTO → minutes
15

Next level
RTO → minutes

RPO → less than a second
15

Based on native Asynchronous
Replication
But easier.... easier is be er !
Data provisioning included (clone)
2 or more nodes
Manual Failover
Transparent/Automatic query routing with
MySQL Router
MySQL InnoDB ReplicaSet
16

17

On the current server:
JS > dba.con gureReplicaSetInstance()
JS > rs=dba.createReplicaSet('myreplicaset')
On a new one where MySQL (server and shell) is just installed:
JS > dba.con gureReplicaSetInstance()
Back on the rst instance:
JS > rs.addInstance('10.0.1.2')
18

MySQL InnoDB ReplicaSet - examples
19

20

21

MySQL Router
It's very easy to con gure MySQL Router with the bootstrap command !
22

MySQL Router
And as usual, visible in MySQL Shell
23

MySQL InnoDB ReplicaSet - Manual Failover
MySQL Shell is connected to MySQL InnoDB ReplicaSet using MySQL Router.
24

MySQL InnoDB ReplicaSet - Manual Failover
25

Next level
Now my service is very important, I would like to be almost always up (automatic failover)
and never loose data !
26

Next level
RTO → seconds
26

Next level
RTO → seconds

RPO → 0
26

Based on native Group Replication
But easier.... easier is still be er !
Data provisioning included (clone)
3 or more nodes (odd number)
Automatic Failover
Uses MySQL Router
Con gurable Consistency Levels
MySQL InnoDB Cluster
27

28

On the ReplicaSet Primary Server:
JS > dba.dropMetadataSchema()
JS > cluster=dba.createCluster('mycluster')
On the ReplicaSet Secondary instance:
SQL > STOP REPLICA;
SQL > RESET ALL REPLICA;
And back on the new Cluster Primary
member:
JS > cluster.addInstance('mysql-2')
And we can add a third note for automatic
fail-over, on that node:
29

On the ReplicaSet Primary Server:
JS > dba.dropMetadataSchema()
JS > cluster=dba.createCluster('mycluster')
On the ReplicaSet Secondary instance:
SQL > STOP REPLICA;
SQL > RESET ALL REPLICA;
And back on the new Cluster Primary
member:
JS > dba.con gureInstance()
From a member of the cluster:
JS > cluster=dba.getCluster()
Finally, don't forget to bootstrap MySQL
Router again:
# mysqlrouter --bootstrap
clusteradmin@single-mysql:3306
--conf-use-gr-noti cations
--user mysqlrouter --force
# systemctl restart mysqlrouter
And we can add a third note for automatic
fail-over, on that node:
29

30

And what's next ?
MySQL InnoDB Cluster with an asynchronous replica for DR:
asynchronous replication source connection failover (8.0.22)
with support for Group Replication (8.0.23)
MySQL InnoDB Cluster with asynchronous Group Replication nodes
and more to come...
31

Questions ?
32

From single MySQL instance to High Availability: the journey to MySQL InnoDB Cluster

More Related Content

What's hot (20)

Similar to From single MySQL instance to High Availability: the journey to MySQL InnoDB Cluster (20)

More from Frederic Descamps (10)

Recently uploaded (20)

From single MySQL instance to High Availability: the journey to MySQL InnoDB Cluster