[gr85z]

I am new to forum but have some VIO background,
Problem discovered yesterday and we are trying to figure out why. SO here is what we have.
2 -p5-570's (seperate frames)
Setup with 1 dev, 1 test, 1 prod LPARS, 2 dev/test VIO servers, 2 prod VIO servers.
Both frames setup exactly the same with VLAN's etc... since we will be introducing HACMP into them later.
Currently we are getting packet loss on the 2 frames (only dev is set up with 2 vio servers). If we shut down the 2 vio servers on one frame the connections are back no issues or loss. When we enable them the packet loss and lack getting logged into them happen again.
So VIO1 has
VEA 1 VLAN 1 Priority 1 (private network 100MB copper)
VEA 2 VLAN 2 Priority 2 (public network 1G Fibre)
VIO2 has
VEA 1 VLAN 1 Priority 2 (private network 100MB copper)
VEA 2 VLAN 2 Priority 1 (public network 1G Fibre)

This is the same for both frames. Any clues to what could be going one these are confugred to the LPAR's as SEA and have a control channel setup with them also.
Could the VLAN's be conflicting with each other on the network?


UPDATE::: We did a lshwres on both frames and the MAC adresses are the same. Anyone have any clue on why it did this?
Thanks for any help

[duke900ssd]

The only time I have seen virtual ethernet mac addresses the same on different systems was after a nightmare install when the IBM CE had been swapping all sorts of bits between machines to try and get the second one to install and boot up properly.

This should never happen because the virtual mac address should be derived from the system serial, the LPAR ID and the adapter ID.

When the CE was swapping bits he must have left the service processor from one machine in the other, so the HMC kind of got confused about its serial number and they both generated the same mac addresses. Same scenario as yours, two systems configured in the same way.

The proper fix was to power them both down and then reset the service processors.

The workaround we used in the mean time was to change the mac address from the HMC with the xxx command, sadly I cannot remember the xxx bit but it was some sort of chdev type command.

Never thought I'd see it again, HTH.

If you had similar install problems, like swapping bits between machines, call IBM, they should be able to sort out the command and proper fix for you.

[gr85z]

Got it fixed.
Ye reset the one frame that wasn't up and running.
Here is what I did
Frame in standby modefrom CLI HMC
chhwres -r virtualio --rsubtype eth -m (managed system name) -o s -a "mac_prefix= (put new hex for first 3 bit)
Then
lpcfgop -m (managed system name) -o clear

Then start rebuilding your LPARS at this point I don't trust the profiles from saved systems or even the same system.

Thanks for reading hope this helps someone in the future before they get to far along in a build.