Re: Aggregate Ethernet

[DrFranken <midrange@xxxxxxxxxxxx>]

On my list of things to try. Have TR3 loaded and available lines, just 
no time to try.

Understand the need for some sort of message somewhere indicating a port 
dropping out, that is critical! If you are using two lines and virtual 
IP you have the visibility of the line failure so that would be a 
preferable solution at this point if monitoring is critical.

I would hope that IPLs DURING a network outage would be a rare bird, 
however point taken that assuring all lines are available when the 
aggregate line is brought up.  Dynamic add of additional ports should be 
enabled too, this is standard stuff on the network side so I'm sure they 
just didn't get that far yet.

 - Larry "DrFranken" Bolhuis

On 11/17/2011 11:41 AM, rob@xxxxxxxxx wrote:
> Anyone trying the new Aggregate Ethernet with 7.1TR3 discussed at
> http://publib.boulder.ibm.com/infocenter/iseries/v7r1m0/topic/rzajy/rzajywhatsnew.htm
>
>
> I am using it.  It has some benefits.  I can drop a port and the line
> still communicates.  Bring that port back active, give it a few seconds
> and drop the other port and it still works.
>
> Needs some work though.
> Item 1:  Let's say you spread one line across two ports.  One port drops.
> You will not see any notice.  Not in QSYSOPR, DSPLOG, STRSST, anywhere.  I
> compare this to losing a drive out of a raid 5 parity set and not getting
> any notice.  IBM may fix this at sometime in the future.  There is
> something on DSPLIND which will show you that - but that's a lot trickier
> to monitor with many of the standard monitoring tools sold for the i (ie
> Messenger Plus and others).
> Item 2:  Let's say one of the ports is down when you vary on the line
> (like you did an IPL while still waiting for the telco or network admin to
> fix the issue).  Ok, so now your line is varied on, with only 1 link
> active.  Line gets fixed.  You cannot bring the second port active without
> varying off/on the whole line.  That's a wee bit too disruptive for my
> tastes.  IBM may fix this at some time in the future.
>
> Rob Berendt