tu0: transmit FIFO underflow

[hy_3]

Digital UNIX 4.0D.There are some messages in the /var/adm/messages:
Jan 16 16:09:24 HOSTNAME vmunix: tu0: transmit FIFO underflow: threshold raised to : 256 bytes
Jan 16 16:09:33 HOSTNAME vmunix: tu0: transmit FIFO underflow: threshold raised to : 512 bytes
What does it mean?Thank you.

[Hein van den Heuvel · Accepted solution]

Before entering a new question please consider doing a quick SEARCH to see if it has been asked and answered already.

This particular question was asked and answered in:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=210880

Basically this is a benign (harmless) warning that the system is doing 'the right thing' by increasing an internal buffer. You can ignore it and be happy, or can can try to pre-increase the buffer before the need is detected. Your end users will not be able to tell the difference eitherway,.

Hein.

[hy_3]

Thank you very much.

[Ralf Puchner]

Hein,

it is not harmless as you described. It will lead normally to a 30 second sync if using switches with spanning tree enabled. During this time the machine is not reachable from outside.

[Joris Denayer]

Hi,

Ralf, this is correct. But it is the Spanning Tree protocol that is learning it's configuration again and is responsible for the timeout.

Solution:
- If the switch allows it, you could disable STP for the port used by the system.
- But, switches like our Procurve's don't allow selective/per port STP settings. However, you can set a per_port_parameter "FAST" that will make the STP calculations much shorter.

[Ralf Puchner]

Joris,

this is quit clear, but the experience is that system administrators doesn't know what was set or if the network hardware is a switch/hub, blaming the software vendor for problems.

So the last posting was an add on to check the parameters on the switch or to configure the driver to prevent further problems.