[mythtv-users] IOBOUND errors over NFS - fixed

[Rudy Zijlstra]

John Andersen wrote:

>On 2/24/06, Rudy Zijlstra <mythtv at edsons.demon.nl> wrote:
>  
>
>>Asher wrote:
>>    
>>
>
>  
>
>>I remain convinced that what the OP did was hiding his true problem. And
>>he will likely run into problem again.
>>    
>>
>
>This seems highly unlikely to me.  These problems typically get solved
>and stay solved.  Clearly he has proven his network can handle this,
>  
>
This is not a question of whether the network can handle it (it 
obviously can) but of network configuration. And configuration in a 
stable way (a way that will work, no matter in which order the equipment 
is powered).

>and it is unlikely it will suddenly stop handling it unless he
>replaces a switch with something less capable.
>  
>
Nope, it may also break if he replaces with something as capable, but 
with a different behaviour.
Actually, in the current configuration it may well break if a different 
power on cycle is used.

>I've seen this solution work many times.
>
>In my case, more than one 3com card has gone directly to the trash bin
>for refusing to properly auto-neg.  Some models garnered an industry
>wide reputation for this.
>  
>
Yup. Please take a look at the time and date of those problems?

>But on several occasions, in servers, the quickest expiedient was
>to lock the card to the setting you wanted, and one which you KNEW
>your network could handle.
>  
>
Again, you also needed to force the *other* side of the link to get 
dependable behaviour. The OP only stated he forced *one* side of the 
link. If you force, force *both* sides, as that is the only dependable 
way with sub 1G equipment.

>The simple truth, inspite of the grandstanding going on in this
>  
>
Thanks for the name-calling ...

>thread, is that auto-neg is iffy in many cases, and can yield dribble
>connections.  Anyone who knows the capabilities of the infrastructure
>is perfectly justified in locking a nic at a setting they have
>demonstrated to work.
>  
>
I've been working with many network components. The older ones clearly 
had some nego problems. On the current this has been solved (as far as i 
have been able to determine). This is not to mean that i like RealTek 
(to name one, as they are not adhering to the PHY level spec of 
ethernet). So there are still enough problems.

Example: one of my links has a tendency to go to 100FDX iso 1000FDX. 
Reason: the cable connector is loose in the frame connector, and may not 
be fully settled....  I need to replace that cable. If the original 
poster has a similar problem (connector not fitting well, retension lip 
broken), such an error may result in auto-nego not working well (and 
thus the NIC going to 100HDX).  If you then force the NIC to 100FDX this 
may work. It may also result in different problems, not easily noted, 
but costing in performance.