Messaging Solutions in a Linux Environment

Chapter 5. Implementing HA queue managers: Part 3 177

Table 5-2 Summary of HA Configuration

5.4 Running the shared-disk test

We are using the same test as in scenario two. Refer to 4.7, “Running the

shared-disk test” on page 136.

In this section, we examine the differences in a failover scenario between the

shared SCSI disk enclosure and shared fiber enclosure setups.

If brk2 node fails, brk1 starts a takeover of its function, that is running the queue

manager qmgr5. See Figure 5-15 on page 178. As both brk1 and brk2 have

simultaneous access to the disks, the file system is already available to brk1 and

no disk takeover is needed as in the scenario with the shared SCSI enclosure.

In a failback situation, when brk2 recovers, the host brk2 first establishes access

to the disks at boot time then takes over the task of running the qmgr5, without the

need to unmount or mount file systems. Because both the nodes have a

connection to the shared disks, both failback and failover situation become

marginally faster than with shared SCSI disks.

Node(s) Up: IP Addresses: Mounted Filesystems:

brk1 and brk2 brk1 - 192.168.10.101

brk2 - 192.168.10.102

brk1 - /var/mqm/qmgrs

brk1 - /var/mqm/log

brk2 - /var/mqm/qmgrs

brk2 - /var/mqm/log

brk1 only brk1 - 192.168.10.101

brk1 - 192.168.10.102

brk1 - /var/mqm/qmgrs

brk1 - /var/mqm/log

brk2 only brk2 - 192.168.10.102

brk2 - 192.168.10.101

brk2 - /var/mqm/qmgrs

brk2 - /var/mqm/log

178 Messaging Solutions in a Linux Environment

Figure 5-15 Scenario three in failover

brk1

qmgr4

brk2

qmgr5

HeartBeat

/var/mqm/qmgrs/qmgr4

/var/mqm/qmgrs/qmgr5

/var/mqm/log/qmgr4

/var/mqm/log/qmgr5