Intel i350 teaming issue

[VKary]

Hello.

Could someone help with the problem below?

The case:

We have set up teaming for 2 network interfaces.

There is huge multicast UDP traffic.

Sometimes we need to join and leave multicast addresses to start\stop listening the data.

After the messages (below) appear in Event log the teaming falls apart.

Event Log:

Current Primary Adapter is switching from: Intel(R) I350 Gigabit Network Connection

Preferred secondary adapter took over: Intel(R) I350 Gigabit Network Connection

Secondary Adapter has rejoined the Team: Intel(R) I350 Gigabit Network Connection

Preferred primary adapter has been detected: Intel(R) I350 Gigabit Network Connection

Current Primary Adapter is switching from: Intel(R) I350 Gigabit Network Connection

Preferred primary adapter took over: Intel(R) I350 Gigabit Network Connection

Little research shows that the problem was initiated by Intel diver in case there are no some heartbeats (even if a heartbeat was late for milliseconds (~100 msec)).

We tried to introduce some delay (2sec) between join\leave multicast group and it allows us to see the problem much less frequently but the problem is still here.

Also we set the settings:

Check Time: 3

Load Balance Refresh Rate: 50

[st4]

Hi VovanKVN,

Thank you for the post. The probe packets are used to check on the status or state of the teamed member in the Team. What you can do is to try disabling the probe packets if you preferred not to have probing packets sent. Please refer to this website for the option http://www.intel.com/support/network/sb/cs-009744.htm Network Connectivity — Advanced Networking Services - Advanced Settings for Teams

You can try use other failover detection method instead of sending probe packets as below:

Link-Based - Monitoring actual link to partner port per adapter.

Activity-Based - Monitoring activity on the wire to determine which has the most traffic.

Thanks,

wb

[VKary]

Hello,

thank you for you answer.

Unfortunately, it does not solve the problem, because we use LACP(IEEE 802.3ad) teaming and there is such a sentence in the article:

"For SLA, 802.3AD teams -> no probes used, probe the settings not present in the Advanced Settings page of team."

[st4]

Hi VovanKVN,

Thank you for the clarification. Let me check on this, can you provide also what is the operating system used? thanks.

rgds,

wb

[VKary]

Hello,

Thank you for the assistance. The operating system is Windows Server 2008 R2.

[VKary]

Hello,

Some additional details:

Intel I350 Interfaces are connected to 2 different Cisco Nexus 5548 switches.

The protocol LACP(IEEE 802.3ad) is used to create a team.

Intel I350 driver: 12.11.97.0 (15.05.2014).

Virtual team adapter version: 9.8.52.0 (25.02.2014).

I can provide you with the system error log if needed.

[st4]

Hi VovanKVN,

Thank you for the information, what is also the brand and model of your server? We will check on this.

rgds,

wb

[VKary]

Hello,

thank you for your assistance. The server is Supermicro SYS-6018U-TRTP+

[st4]

Hi VovanKVN,

Thank you for the information, will update you if there is any finding.

rgds,

wb

[VKary]

Hello Support team,

Could you please let me know if you have any updates on this?

[st4]

Hi VovanKVN,

Thank you for the ffollow up. we are still checking on this. Will update you if there is any finding.

rgds,

wb

[st4]

Hi VovanKVN,

Good day. We need additional info to futher check:

1. how many interfaces are members of a team,

2. How are the NICs connected to the switches?

thanks,

wb

[VKary]

Hello,

Sorry for the delay.

The additional information:

1. two interfaces.
2. NICs connected to the switches by UTP 6 patch cord(RJ45).

Would like me to provide you with some other details?

[st4]

Hi VovanKVN,

Thank you for the information. I will further check and get additional info if needed.

rgds,

wb

[st4]

Hi VovanKVN,

Good day. Please find below hope can clarify:

In 802.3ad teaming mode probes are not used. But it used LACP protocol to communicate between switch and a team. LACP frames are

multicasts frames. So it is OK to see multicast frames sent to 01:80:C2:00:00:02. If there are too many such multicasts and there is

problem with aggregated traffic then there is some other configuration issue here.The possible problem here might be there are two team members connected

to two different switches. It is not possible to aggregate team members in one aggregator when they are connected to different switches.

Thanks,

wb

[VKary]

Hello.

Thank you a lot for the information