LAN Configuration and Management Manual

Introduction to the SLSA Subsystem

LAN Configuration and Management Manual—520469-012

2-16

Fault Tolerance of the SLSA Subsystem

The SLSA subsystem operates even after a loss of one of its components. This section

describes how the SLSA subsystem maintains data paths to the LAN subsystem when

a processor, SAC, media, or adapter is unavailable.

Loss of a ServerNet Fabric

When access through a ServerNet fabric is lost, all traffic goes through the remaining

fabric. When the failing fabric is restored, all traffic goes over both fabrics again. If both

fabrics fail, traffic moves as described in Loss of Access of One Processor to a SAC on

page 2-16 or Loss of Access of All Processors to a SAC on page 2-17.

When access through a ServerNet fabric is lost, SLSA events are generated and

stored in the Event Management Service (EMS) log. Transient-fault event messages

that show a transient-fault number of 3044 (Path switched to X fabric) or 3045 (Path

switched to Y fabric) may indicate that access across a fabric was lost.

Loss of Access of One Processor to a SAC

If a SAC on an adapter becomes unavailable to a LANMON that has a data path to the

SAC, all clients (such as TCP/IP, PAM, and IPX/SPX) on that processor lose access,

and the SLSA subsystem performs the following steps:

1. The LANMON detects that it cannot communicate with the SAC because:

a. The SAC issued a command that timed out.

b. The SAC received an error.

c. The SAC did not respond after 3 consecutive pings sent 10 seconds apart.

2. The LANMON tries to recover access to the SAC by trying the alternate ServerNet

fabric. If the LANMON does not succeed:

a. The LANMON informs the LANMAN process about lost access to the SAC.

b. If the LANMON that lost access was the owning LANMON, the LANMAN

process assigns ownership of the SAC to another processor, but only if the

LANMON in that processor can access the SAC.

c. LANMAN checks whether any other processors have access, and, if none

does, proceeds to the behavior described in Loss of Access of All Processors

to a SAC on page 2-17. If other processors have access to the SAC, LANMAN

repeats Step 2b (above).

If access is not reestablished, the clients have to wait until the SAC automatically

comes up again.