Linux clusters-Principles of highly available clusters

HA:

Resource stickiness

Resource constraints

The arrangement constrains the operation to be able to

Whether a resource can run on the same node

Score

Positive values: can be together

Negative value: cannot be together

Location constraints

Positive values tend to this node

Negative values tend to escape this node

Sequential constraints

Define the order in which resources are started or closed

Vip,ipvs

Start Order IPVS-->VIP

Close Order Vip-->ipvs

-inf: Negative Infinity

INF: Positive Infinity

Resource Isolation

Node Level: STONITH

Resource level:

Columns such as: FC SAN switch

STONITH:

Split-brain: Cluster node can not effectively obtain the other side of the heartbeat information, creating a brain fissure

One consequence: preemption of shared storage

Active/active High-availability clusters

IDE (ATA) 130M

SATA 600M

7200rpm

iops:100

SCSI 320M

Sas

15000rpm

IOPS 200

USB 3.0 400M

Machinery:

Random Read and Write

Sequential Read and write

Solid:

Ide,scsi and the same mouth

SATA, SAS,USB serial port

Das

Direct attached Storage

Directly to the native bus, bus

Files: Blocks

Nas

Network

File server: File level

SAN

Storage Area Network

Storage Area Network

Scsi:smail Computer System Interface

Heartbeat v1 comes with a resource manager

Haresources

Heartbeat v2 comes with a resource manager

Haresources

Crm

Heartbeat V3 Explorer CRM developed as a standalone project, pacemaker

Resource

Primitive

Clone

Group

Master/slave

RA Classes:

Legacy Heartbeat v1 RA

LSB (/etc/rc.d/init.d/)

OCF (Open Cluster Framework)

Pacemaker

Linbit (DRBD)

STONITH

Resource Isolation Level

Node level

Stontih

Resource level

FC SAN Switch

STONITH Equipment

1. Power Switch PDU

2.UPS

3. Blade Server

4.lights-out Devices

5.Testing Devices

Meatware

Heartbeat

udp:694 Port

This article from "Operation and maintenance Growth Road" blog, declined reprint!

Linux clusters-Principles of highly available clusters