Network cards - specs / disfunction

[ieronymous]

Hi

A generalize title since it contains different network cards and issues Im dealing with.
I ll start with the informative one which I need your confirmation.
I have installed in several servers the 10Gbe Mellanox Connect-x 3 pro
(X3 R740 / X2 R730 / X1 T7610) card which consists of 2 ports and in all of them running lspci gives weird output.
Most ofthe cards are HP branded but the same goes for the one from Mellanox.

lspci | grep Ethernet
01:00.0 Ethernet controller: Intel Corporation I350 Gigabit Network Connection (rev 01)
01:00.1 Ethernet controller: Intel Corporation I350 Gigabit Network Connection (rev 01)
18:00.0 Ethernet controller: Intel Corporation 82599ES 10-Gigabit SFI/SFP+ Network Connection (rev 01)
18:00.1 Ethernet controller: Intel Corporation 82599ES 10-Gigabit SFI/SFP+ Network Connection (rev 01)
3b:00.0 Ethernet controller: Mellanox Technologies MT27520 Family [ConnectX-3 Pro]

Notice that Mellanox Technologies MT27520 Family [ConnectX-3 Pro] appears only one time, even though the card has two ports.
Is this because of the iommu properties of the specific slot (already changed to a different slots-same outcome) or motherboard or card?
Seems weird since in daughterboard card Intel Corporation 82599ES 10-Gigabit SFI/SFP+ Network Connection (rev 01) appears 2 times as it should.
Am I missing something here?

On the same topic different card, an Intel Corporation I350 Gigabit Network Connection (rev 01) with 4 ports but in case it matters a CNA card,
has been configured as below
vmbr1 based on bond1 based on the 2 first ports of the card in LACP 802.3ad mode
vmbr2 based on bond2 based on the 2 other ports of the card in LACP 802.3ad mode
I accidentally noticed that when pinging with -t parameter (so continuously), vmbr1 looses at least one ping before reaches 30 sec.
That number each time is different, but not even once managed to transfer all ping requests successfully.
When I transfer the management to vmbr2 (instead of vmbr1 was previously), no packet loss while sending icmp for more than 1 minute.

new edit: Changing mode from LACP 802.3ad to balance-rr in bond1, seems to make the ping go on continuously . Weird thing is the other 2 ports in LACP mode the play along well. So it isn t a support feature issue. All ports are on the same card.

Does this card considered to be bugged / malfunctioned / problematic?

Thank you in advance.

 

[ieronymous]

I probably found the problem with my second issue.

Since in the server room there are over 400 eth cables, I thought I was using the right ones for LACP mode but it turns out I didnt.
By correct cables I mean the ones terminated to the appropriate ports of the switch which has those ports configured in LACP mode as well.
Button line, used wrong set of cables.

Any insight though for the Mellanox card?

 

[Pifouney]

Hey

Haven't chirugical idea about your ethernet card Mellanox.

But, for deeply test of this hardware, a crashtest solution can be:

Create a VM(use Os of your choice)
Before boot them, PCI Passthrough the medllanox NIC (comment her in interfaces file PRIORLY)
See how she's detected in you VM. Maybe a specific procedure needed for spawn your mellanox adapters (screenshot seem to only see the melanox controller)

 

[ieronymous]

See how she's detected in you VM. Maybe a specific procedure needed for spawn your mellanox adapters (screenshot seem to only see the melanox controller)

Thank you for your interest. The way its going to show up in OS depends on how I shall assign the card to the system beforehand.
What I mean by that is the options I need to tick like multifunction, rombar...etc which state that the card has 2 ports. Its the same as gpu where you have to pass the audio part too since it is a separate chip.

Yet this would mean anything to my initial question. I d like to know if others having this card appear in cli like mine does.