C3000 Enclosure and two Cisco 3020 switches

[Antonio Francucci]

Hi to all, at a costumer site we got an HP C3000 Eclosure with two Cisco 3020 Blade Switches in I/O bay 1 and bay 2.

To test the redundancy we have connected to each 3020 an uplink copper cable to a core switch. We have then installed a Red Hat distro in one of the BL460c G6 blade in the C3000 (each blade has the two onboard nic and a quad-port mezzanine in mezzanine slot1).
Red Hat sees 6 nics (two of them disconnected because we have a 4 port mezzanine in slot1 of the blades and in that slot we can use only two ports).

We have then launched 4 pings to each nic ip from a remote pc connected to the same core switch of the two uplinks form the 3020.

At that point we have disconnected the uplink cable for the 3020 in I/O bay 1 of the C3000 and the two nics connected to that switch (the two onboard of the BL460c G6) stopped responding to ping as expected. Reconnectnig the uplink resumed the ping responses.

The problem came when we disconnect the uplink cable form the second 3020 in I/O bay 2 of the C3000, in fact after we pulled the cable ALL of the four pings stopped responding...
The same happens if we power off the second 3020....

We know that the two 3020 are not linked togheter but why they act in this way ?

Cross connecting the two 3020 (via port 23 and 24 internally) would be a solution?

Can we suggest to our costumer to set up an etherchannel?

Thanks in advance!

[Antonio Milanese]

Hi Antonio =)

first there's something wrong on your reported configuration since AFAIK half-height blades does not support quad card in mezz 1:

LOM (1,2) -> IC1
MEZZ1(1,2) -> IC2
MEZZ2(1,3) -> IC3
MEZZ2(2,4) -> IC4

http://h20000.www2.hp.com/bizsupport/TechSupport/CoreRedirect.jsp?redirectReason=DocIndexPDF&prodSeriesId=3540808&targetPage=http%3A%2F%2Fbizsupport2.austin.hp.com%2Fbc%2Fdocs%2Fsupport%2FSupportManual%2Fc01508406%2Fc01508406.pdf

Anyway with multiple c3020 blades and multiple nics I would like to:

1) xconnect the two 3020
2) configure (M)STP on c3020 and upstreams switchs to avoid loops
3) configure link state traking groups to reflect path failures to downstream blades ports
2) bond host nics with policy 1 (NFT) or policy 6 (ALB)

Regards,

Antonio

[gregersenj]

I will recommed to read the C3000 Technology brief.

You will find it on the web.

BR
/jag

[Antonio Francucci]

Thank you for the responses (i will award points :) )

i've tried to reset the two 3020 to the default configuration and then to configure link state tracking and cross connect the switches with no success, i suspect that the NC364m quad port mezzanine in slot 1 of the BL460c G6 is not fully compatible and i want to change it with a dual port mezzanine instead (an NC360m).

What do you think about that?

[Antonio Milanese]

Hi Antonio,

yes..changing the mezzanine with a dual port is the way to go!

I understand that you have the ability to update "offline" so it's a good idea to align the firmware version to latest supported ones

http://h18004.www1.hp.com/products/blades/bladesystemupdatestable.html

especially for the C3200 (12.2.55 IIRC) there are some MSTP and LACP fixes when upstream switches are non cisco ones

For MSTP pay attention to use a common config/revision and VLAN mapping then set path cost accordingly to load balance traffic and avoiding unoptimal use of xconnect to left the enc..and last but not least set STP prio so MIST and CIST root are elected outside the enclosure.

Regards,

Antonio