3.1.2 Matrix Server RHEL4 Release Notes

PolyServe Matrix Server Release Notes 24

5814 QLogic driver can exhaust kernel memory

Under certain high-stress I/O conditions, the QLogic

FibreChannel driver can exhaust kernel memory.

This condition typically occurs while a server is running at very

high I/O rates, and is possibly triggered by storage subsystem

error recovery. During this period the system administrator may

see many messages of the form:

Apr 30 10:06:47 qar54s1 kernel: scsi(1:5:12): QUEUE FULL status

detected 0x1c-0x928, pid=128300970.

Apr 30 10:06:47 qar54s1 kernel: scsi(1:5:4): QUEUE FULL status

detected 0x1c-0x928, pid=128300892.

Kernel memory exhaustion can result in the failure of

applications or ejection of the server from the matrix. If this

condition is encountered, it is recommended that the command

tag queue depth be reduced from the default of 32 to 16 or less.

7435 QLogic Switch login can cause Matrix Server failures

Matrix Server will not start if you are logged into a QLogic

FibreChannel switch and “admin start” is set. A message such

as the following will appear:

Switch address is not responding to SNMP SET requests.

Verify the configured community string has SNMP write

privileges.

The FibreChannel switches are not responding to SNMP requests.

Verify that the switches are configured to accept SNMP access

from this server using the configured community string. The

FibreChannel switches and Matrix Server must be configured

with the same community string.

If you log into the switch with “admin start” while Matrix

Server is running, this switch problem can affect fencing

operations and may cause the matrix to hang.

Workaround. Avoid putting the switch into “admin start” mode

while Matrix Server is running. If it is necessary to use this

mode while Matrix Server is running, limit the use of the mode

as much as possible.

Defect

Description