AW: interface checking for heartbeat

[Alan Robertson]

Jeroen Vriesman wrote:
> 
> Isn't pinging  other hosts a bit of a quick and dirty way to check the NIC?
> 
> Looking in e.g. tulip.c I see lots of ways to check if the link is up e.g. the Tx
> buffer getting flushed is a sign there is something wrong with cable/nic/hub.
> Would be very device dependend.
> 
> I think I would use (in the case of the tulip driver):
> 
> printk(KERN_WARNING "%s: Tx hung, %d vs. %d.\n",
>                            dev->name, tp->cur_tx, tp->dirty_tx);
>                 tulip_tx_timeout(dev);
> 
> >From tulip.c
> Not so hard to implement, just put the messages of the nic's in a seperate
> logfile and let heartbeat check that log now and then.
> 
> Would this be a good idea?
> 


Monitoring logs is always a good idea.  Running diagnostic tests proactively
is also a good idea.

	-- Alan Robertson
	   alanr@suse.com