----- Original Message -----
From: Stuart Henderson <[email protected]>
To: Bogdan <[email protected]>
Cc: "[email protected]" <[email protected]>
Sent: Wednesday, August 10, 2011 6:33 PM
Subject: Re: interface trunk0

On 2011/08/10 08:01, Bogdan wrote:
> 
> The information from ifconfig -A and maybe also netstat -rn are
> pretty important in showing what is wrong. I have an idea what you
> missed, but
> I'm not going to guess based on incomplete information.
> 
> This would require to
> commute back to trunk setup, which I can not do it anymore
> but I thank you for
> the willing to help me.

Maybe you could do this on a test machine or VM..

I have no machine available to me.

If you want to let me know what the solution is would be greate.

I forgot to mention that I also flushed the routing table and added the default 
gateway
bounded to trunk0 rather than em0, but that also failed to work.

> It seemed pretty secure to activate
> trunk, although had no experience with that in the past.

I had much less trouble with trunk on OpenBSD than with
bonding on Linux. But personally I would try to avoid making
this type of change in production without a redundant machine
and out-of-band console access.

I do not consider linux+bonding a solution as I would never 'trade' on OpenBSD 
box on otherOS.


>> 4.0 is nearly 5 years old - how are your
>> services critical for
>> local infrastructure going to cope when the hardware
>> fails?
> 
> well, good question.
> 
> First I took the courage to re-write from
> scratch all the software 
> with concurrency, distribution, fault tolerance and
> replication in mind,
> 
> Many services are self-replicating on secondary machine,
> so I achieve a decent level of data persistence.
> 
> All the code I wrote is also
> commited into a CVS repository which is mirrored on several machines,
> including the development machine.
> 
> I don't know how good this setup is but is
> something.

This is all good, and much better than many people will do.

> And yes a binary image of the running OS would be great, but I am
> keep telling myself that nothing wrong could happen.

A binary image might not be so good; if you need to replace with
new hardware, you may well find that it requires an OS with newer
NIC drivers, and/or supporting ACPI for correct interrupt routing,
so a direct clone of the 4.0 system might not work, forcing an
upgrade. IMO it's much better to do this as planned maintenance
(with a good back-out plan in case of problems) rather than wait
until forced. 

This is better than OS imaging

Reply via email to