6-2 Sun Fire™ B1600 Chassis and B100s, B100x, and B200x Blade Product Notes • May 2004
6.1 Switch Firmware Issues
The following known problems apply to the current release of the switch firmware
for this product:
4899178: Blade network traffic is only allowed through the IP filter on the VLAN
configured as the management VLAN.
The management VLAN is the VLAN that has been assigned an IP address to
allow network access to the switch’s management interfaces (by default this is
VLAN2). Other VLANs can be assigned to the NETMGT port to allow blades to
talk to particular hosts on the management network. The usual way to do this
would be to put particular blades and particular hosts that are on the
management network onto a tagged VLAN that is separate from the management
VLAN. However,the switch’s packet filter will not forward any traffic from the
blades to the management network unless that traffic is for the management
VLAN. This is a problem that will be fixed in the next release of the switch
firmware. It means that trafficfrom the blades will not be seen by hosts on the
management network that are external to the chassis (in other words, only other
blades in the chassis will see the traffic) unless the blade, the switch, and the
external hosts involved are all on the management VLAN. Note that the
configurations for multiple tenants described in Chapter 7 of the Sun Fire B1600
Blade System Chassis Software Setup Guide will not be possible until this
problem has been fixed.
4854587: It is possible that the System Controller will reset the switch when the
switch is executing commands that requireunusually intensive processing: The
SC continuously polls the switch for status as part of the system healthcheck.Itis
theoretically possible that, while executing commands requiring unusually
intensive processing, the switch will be unable to respond to the SC’s status
request within the timeout period because it must first complete execution of a
process-intensive command.
This should not happen during normal operation. The problem was observed
when a user sent a sequence of lengthy commands (for example, commands
adding many VLANs to a port) to the switch without waiting for the prompt
between each command. This filled the switch’s input buffer and blocked the
status poll messages.
Toavoid the problem, always wait for one command to complete before issuing
another command on the CLI. If you are using scripts this is especially important.