C7000 enclosure with Flex-10 Networking

[Eric Sabo_1]

We been working with HP Support but haven't been able to get anyone to answer the following questions about a C7000 Enclosure using the flex-10 connector in Bays 1 and Bay 2 (BL465 blades fully populated). We have a mixture of VMs and physical servers running in this environment.

In the management port logging we are seeing "Network packet flooding detected" - would this impact the flex 10 connectors? Or are these two network connectors totally separate? What is the optimal settings for speed/duplex on the management port - Does the Cisco switch have to match settings?

We do see from time to time some request timeout during pings back to the host. This is connected via fiber back to a Cisco 6513 - which show no packet loss.

Any help would be appreciated?

[HEM_2]

The 'network packet flooding detected' message comes from the OA syslog and means that you are seeing excessive traffic on the OA Management Network, i.e. where the OA/ILO port connects.

This could affect connectivity to the OA, ILO's or the management interface of Virtual Connect Manager.

This doesn't mean that network packet flooding is occurring where the VC uplinks connect to your network.

The bladesystem management path is architected completely separate from the data path.

Now, if (and only if) your OA/ILO port connects to the same VLAN as what exists on your VC uplinks then it is possible that the network packet flooding that the OA is detecting on its network is severe enough to impact the VC data path (because they are on the same VLAN)

best setting for OA/ILO port and the switch it connects to is Auto/Auto. Settings need to match on both sides or a duplex mismatch could occur...

hope this helps

[Eric Sabo_1]

Yes, our OA and the VC are in the same VLAN, so you are saying we should place different VLANs to prevent this conflict. Do you know if there is any documentation on this?

[HEM_2]

Eric,

my recommendation would be to run a protocol analyzer on your VLAN and check for excessive multicast/broadcast or unicast flooding. If you see something unexpected, look for the source and correct that which should improve network performance overall.

If the traffic you see is expected/normal for your environment than the OA message may just be a little too sensitive.

not sure where this might be documented... maybe in the OA user guide?

[rick jones]

If you have a 1Gbit VLAN and the OA is but 100Mbit, and you have a lot of broadcast and perhaps multicast traffic, or so many MACs in the VLAN that the forwarding tables in the switch to which the OA is connected are full, then you can quite possibly overload the poor OA.

If you also see ping drops back to the host IPs (ie the LOMs etc on the blades themselves) that suggests you have a rather busy network indeed and you may want to start diagnosing that - you may have some number of nodes spewing packets into the network...