Topic
5 replies Latest Post - ‏2012-07-26T13:00:46Z by MarkRow
MarkRow
MarkRow
3 Posts
ACCEPTED ANSWER

Pinned topic Failing SEA - what have I missed?

‏2012-04-03T16:25:34Z |
I have 2 VIO servers in a 9117-MMC (P7-770) bridging 5 networks on the ETHERNET0 switch.
The first 4 networks have the SEA using an Etherchannel as the "Physical" device and they work without problems.
The fifth network is using a single physical connection (a 1Gbps port from the Integrated MultiFunction Card)on each VIO server, and this is the connection I have a problem with.
There is no VLAN tagging on any of the networks and there are a number of different switches these NICs are connected to.
All 5 of these networks use SEA failover.

The VIO servers have an IP address configured on the SEA and they can contact external servers and the servers within the 770. The VIO Clients can contact each other and the VIO servers but nothing else.

When an external server attempts to contact a VIO client they get a complete entry (with MAC) in their ARP cache for the VIO client and the VIO client gets a complete entry for the external server.
When the VIO client attempts to contact the external server it gets an (incomplete) entry in the ARP cache for the external server but the external server gets a complete entry for the VIO client.

I had tried a configuration without SEA failover, using Etherchannel NIB on the VIO client, but this failed as well.

Is this likely to be an SEA configuration error or a problem with the physical switches the network cards are connected to for this network?
Any suggestions as to what I can look at next to resolve this, my network support team say the switch configurations are all the same?

The seastat command shows activity for my VIO Client (by MAC address) but doesn't show an IP address so do I have an ARP problem and what could be the cause?
Two of the three physical switches that manage the network are using a default spanning tree priority and I was wondering whether the Virtual switch may operate in conflict with them in this case (although the third does have a higher priority - 8192). The primary VIO Server is connected to one with default priority and the secondary to the one with the higher priority. If I put the primary into standby it doesn't make any difference.

I've ruled out the Integrated MultiFunction Card as the problem as I have another network working from another port in the card using a different virtual switch. This is setup in the non SEA failover configuration that failed with my problem network.

Any advice or pointers would be very welcome.

...Mark Rowlinson
Updated on 2012-07-26T13:00:46Z at 2012-07-26T13:00:46Z by MarkRow
  • j.gann
    j.gann
    52 Posts
    ACCEPTED ANSWER

    Re: Failing SEA - what have I missed?

    ‏2012-04-04T16:57:08Z  in response to MarkRow
    by "integrated multifunction card" you probably mean IVE (integrated virtual ethernet)? Be sure you have IVE port parameters set correctly via the HMC. For use in a SEA the IVE port must be set to "promiscuous mode". Read the IVE redbook if needed.

    Apart from this I'd start troubleshooting on a low level. Leaving all SEA aside for a start, do you have connectivity when simply configuring an IP on the ive port in question's interface ?
    • MarkRow
      MarkRow
      3 Posts
      ACCEPTED ANSWER

      Re: Failing SEA - what have I missed?

      ‏2012-04-04T17:16:59Z  in response to j.gann
      The IVE was a feature for the 9117-MMB but has now been removed. The 9117-MMC has the new integrated Multi Function Card which just supplies four additional network ports that can only be be assigned to a single LPAR. Having IVE would be far simpler as I'd just assign the ports to all my LPARs and wouldn't use the VIO server to provide this network.

      I do have outbound connectivity from the VIO server from it's IP address configured on the SEA and when I was first checking the patching I was able to access the network over these ports.

      It's the SEA bridging function that is failing but there is no indicator as to why. It is also only failing for one of my IMFC ports so I'm thinking it's something on the switched I'm connected to but I don't know what to look at, or get my network support team to look at.

      ...Mark R
      • seroyer
        seroyer
        352 Posts
        ACCEPTED ANSWER

        Re: Failing SEA - what have I missed?

        ‏2012-04-04T17:23:49Z  in response to MarkRow
        I'm not sure what would cause that behaviour, but I could see it being a switch port setting. Possibly something to do with seeing traffic from multiple MAC addresses on the same port? I think I'd recommend contacting your support rep.

        Steve
      • j.gann
        j.gann
        52 Posts
        ACCEPTED ANSWER

        Re: Failing SEA - what have I missed?

        ‏2012-04-04T17:26:10Z  in response to MarkRow
        right about ive, I've overread that its a C-Model.

        from a troubleshooting perspective: swapping the cables connecting the switches to the imfc ports should give a clue. does the symptom stay on the same port? does it "move"?
  • MarkRow
    MarkRow
    3 Posts
    ACCEPTED ANSWER

    Re: Failing SEA - what have I missed?

    ‏2012-07-26T13:00:46Z  in response to MarkRow
    Turns out this was a device driver problem that was resolved in VIOS 2.2.1.4 FP25.

    I initially verified this by just updating the device drivers IBM support pointed me at and have just completed updating the servers to the full 2.2.1.4 level.

    As this level is based on AIX 6.1 TL7 SP4 I had to update my NIM server first so I can restore future VIOS backups, if needed.

    ...Mark