Você está na página 1de 26

Veritas Cluster Server 6.

0 for UNIX:
Cluster Management
Lesson 3: Startup and Failover Policies
1

Lesson introduction
Lesson 1: Service Group Dependencies
Lesson 2: Reconfiguring Cluster Membership
Lesson 3: Startup and Failover Policies
Lesson 4: Alternate Network Configurations
Lesson 5: High Availability in the Enterprise

Lesson objectives
Topic

Objectives

Startup rules and


policies

Describe the rules and policies for service


group startup.

Failover rules and


policies

Describe the rules and policies for service


group failover.

Limits and
prerequisites

Use Limits and Prerequisites to control


service group startup and failover.

Modeling startup and Use the Simulator to model startup and


failover behavior
failover policies.

Startup rules
and policies
After completing this topic, you
will be able to describe rules and
policies controlling service group
startup.

Rules for automatic service group startup


Occurs when hastart is executed.
Service group attributes must be set:
AutoStart = 1 (default)
Nonempty AutoStartList
Not Frozen

All systems in SystemList with GAB membership (port


a) must be up and running.
All resources must be probed on systems in
SystemList with GAB membership.
Service group dependencies to child service groups
must be met.
5

Startup system selection


1
1

2
1

The initial set of candidate systems is derived from the


AutoStartList attribute.
Systems are eliminated if:
They are frozen.
The service group has a FAULTED status.
They do not meet service group requirements.

The target system is then selected based on the startup policy of


the service group.

Automatic startup policies

AutoStartPolicy
attribute values

Order

System selection basis

Ordered listed, left to right, in


AutoStartList
(default)

Priority

Lowest priority number in SystemList

Load

Greatest available capacity

hagrp modify dbsg AutoStartPolicy Priority

AutoStartPolicy=Order
The first available system
in AutoStartList is selected.

Group A (
SystemList = { s1=2, s2=1, s3=0 }
AutoStartList = { s1, s2, s3 }
)
Group B (
SystemList = { s1=0, s2=1 }
AutoStartList = { s2, s1 }
)

s1

s2

s3

AutoStartPolicy=Order is
default. It will not appear in
main.cf

Group C (
SystemList = { s1=3, s2=1, s3=2 }
AutoStartList = { s3, s1, s2 }
)

AutoStartPolicy=Priority
The lowest-numbered
system in SystemList that
is also in AutoStartList is
selected.

Group A (
SystemList = { S1=2, S2=1, S3=0 }
AutoStartList = { S1, S2, S3 }
AutoStartPolicy = Priority
)
Group B (
SystemList = { S1=0, S2=1 }
AutoStartList = { S2, S1 }
AutoStartPolicy = Priority
)

S1

S2

S3
A

B
C

Group C (
SystemList = { S1=1, S2=2, S3=3 }
AutoStartList = { S3, S1, S2 }
AutoStartPolicy = Priority
)

AutoStartPolicy=Load
s3

s2

s1

Before C starts, s2 has


AvailableCapacity = 200

A
B
Capacity

300

200

Load

150

75

= Available
Capacity

150

125

100

The system with the


greatest
AvailableCapacity is
selected.

100

Group C (
SystemList = { s1=1, s2=2, s3=3 }
AutoStartList = { s3, s1, s2 }
AutoStartPolicy = Load
Load = 75
)

10

Failover rules
and policies
After completing this topic, you
will be able to describe rules and
policies controlling service group
failover.

11

Rules for automatic service group failover


The service group must have a critical resource
(Critical=1).
These service group attributes must be set to default
values:
AutoFailOver = 1
ManageFaults = ALL

The service group cannot be frozen.


At least one system in the service groups SystemList
attribute must be up and running.

12

Failover system selection


1
1

The initial set of candidate systems is derived from the


SystemList attribute.
Systems are eliminated if:
They are frozen.
The service group has a FAULTED status.
They do not meet service group requirements.

3
1

The target system is then selected based on the failover policy of


the service group.

13

Failover policies
FailOverPolicy Attribute
Values

Priority

System Selection Basis

Lowest priority number in SystemList


(default)

RoundRobin

Least number of active service


groups

Load

Greatest available capacity

hagrp modify dbsg FailOverPolicy RoundRobin

14

FailOverPolicy=Priority
The lowest- numbered
system in SystemList is
selected.
s3

s2

s1

C
FailOverPolicy=Priority is
default. It will not appear in
main.cf

Group C (
SystemList = { s3=0, s1=1, s2=2 }
AutoStartList = { s3, s1, s2 }
)

15

FailOverPolicy=RoundRobin
The system with the
fewest running service
groups is selected.
s3

s2

s1

Group C (
SystemList = { s3=0, s1=1, s2=2 }
AutoStartList = { s3, s1, s2 }
FailOverPolicy = RoundRobin
)

16

FailOverPolicy=Load
s2

s1

s4

s3
B

Before failover

C
Capacity
Load
= Available
Capacity

300

200

100

75

75

25

225

125

75

The system with the


greatest
AvailableCapacity is
selected.

Group C (
SystemList = { s3=0, s1=1, s2=2, s4=3 }
AutoStartList = { s3, s1, s2, s4 }
FailOverPolicy = Load
Load = 75
)

17

Configuring Load and Capacity


hasys modify s1 Capacity 300

main.cf

System s1(
Capacity = 300
. . .
)

hagrp modify dbsg Load 75


group dbsg(
SystemList = { s1 = 1, s2 = 2, s3 = 3 }
. . .
Load = 75
)

18

Limits and
Prerequisites
After completing this topic, you
will be able to use Limits and
Prerequisites to further control
service group startup and failover.

19

Startup example
s2

s1

s4

s3

E
DBs

DBs

DBs

DBs

Limits

Prerequisites

= CurrentLimits

Group E (
SystemList = { s1=0, s2=1, s3=2, s4=3 }
AutoStartList = { s1, s2, s3, s4 }
Prerequisites = { DBs=1 }
)

Before E Starts

20

Failover example
s2

s1

s4

s3
c

E
DBs

DBs

DBs

DBs

Limits

Prerequisites

= CurrentLimits

Group C cannot fail over


C

Group C (
SystemList = { s3=0, s1=1, s2=2, s4=3 }
AutoStartList = { s3, s1, s2, s4 }
Prerequisites = { DBs=1 }
)

Limits are hard


versus Load
which is soft
(allows
negative
values).

21

Configuring Limits and Prerequisites


hasys modify s1 Limits DBs 2

main.cf

System s2(
Limits = { DBs = 2 }
. . .
)

hagrp modify dbsg Prerequisites DBs 1


group dbsg(
. . .
Prerequisites = { DBs = 1 }
)

22

Modeling startup and


failover behavior
After completing this topic, you
will be able to model startup and
failover policies.

23

Using the Simulator


To test startup and
failover policies in the
Simulator:
1. Create a new Simulated
cluster.
2. Copy the main.cf file
from your cluster into the
new configuration
directory.

24

Lesson summary
Key points
Startup and failover policies provide fine-grained control of
service group startup and failover.
You can use the Simulator to model behavior before you
implement policies in the cluster.

Reference materials
Veritas Cluster Server Users Guide

25

End of Presentation

26

Você também pode gostar