High Availability Monitors Version A.03.20 Release Notes

12 Chapter1

Patches and Fixes in this Version

JAGab82383 When a disk resource is a ServiceGuard dependency, the

package may fail over to an alternate node if the user tries to

add or remove volumes from a VG while the package is

running.

Also, the syslog ﬁle hsa errors like this one:

libsgcl.2[nnn]: Couldn’t open /etc/lvmtab

(DRM-40)

The problem occurs because LVM is in the middle of updating

/etc/lvmtab and has the ﬁle locked. Periodically, diskmond

checks the ﬁle for changes. If it is locked, diskmond cannot

complete a successful open and/or read of the ﬁle.

In Version A.03.20, EMS modiﬁed the Disk Monitor so that

the /etc/lvmtab can be opened and read successfully.

JAGad00041 The client did not receive a subclass reply from the registrar

because of a timeout. The timeout occurred because the HA

Disk Monitor diskmond was unable to complete a

“test_unit_ready” when querying a new type of disk that was

active on this system. The new device was an HP SureStore

Fibre Channel Model 60.

In Version A.03.20, the Disk Monitor “test_unit_ready” logic

now clears extraneous bits in the command buffer that are

only needed if performing a “read” operation.

JAGab69421 ServiceGuard package status is set to UNKNOWN if host

aliases are used.

The package monitor was examining all the aliases returned

by gethostbyname. However, ServiceGuard only uses the

ofﬁcial host name. Therefore, the package monitor was unable

to determine where the package was executing.

In Version A.03.20, the package monitor was modiﬁed to use

the same name that ServiceGuard does..

Table 1-3 Corrected Defects

Defect number Problem and Resolution