Aruba & ProVision-based
cancel
Showing results for 
Search instead for 
Did you mean: 

2510G-24 problem - sends some unicast traffic to unused (mac free) ports

 
Highlighted
vlsetka
Occasional Visitor

2510G-24 problem - sends some unicast traffic to unused (mac free) ports

Hi, I have some problem with 2 pieces of 2510G-24 procurve switches. Scenario:

- 2 × 2510G-24, names SW-A01, SW-A02; same config (except ip and name) - attached

- firmware updated to the latest Y.11.35

- 1 × HP server, 4 gb ethernet ports, 2 ports connected to SW-A01 (if 3, 4), 2 ports connected to SW-A02 (if 3, 4)

- this 2 switches are interconnected using LACP trunk on port 21,22

- server is running the Debian linux, all 4 ports are grouped to "balance-alb" mode bonding interface, several VLANs above it

- some ports of SW-A01 are used as tagged uplinks to access switches

 

All works OK, traffic from server is well balanced across it's 4 ethernets, connectivity to all VLANs is OK.

 

There is an border firewall connected to port 1 on SW-A01, and outgoing traffic about 3 Mb/s flows there on VLAN 99. OK.

 

NOW THE WEIRD THING: When I connect some device (may be unmanaged switch, may be an inactive interface on some router) to SW-A02, port 1, I see outgoing traffic on this port - it's something between 0.5 - 2 Mb/s.  Analysis showed me, that it's mainly unicast traffic - part of traffic on VLAN 99. But this port has no MAC active on this port at all:

 

sw-a02 (HP 2510G-24)# show mac-address 1

Status and Counters - Port Address Table - 1

MAC Address
-------------

 

The same is for port 2 on SW-A02, which is configured the same way.

 

SO, as a basic thing: No MAC on this port, but noticeable outgoing unicast traffic:

 

sw-a02 (HP 2510G-24)# show interfaces 1

Status and Counters - Port Counters for port 1

Name :
Link Status : Up
Totals (Since boot or last clear) :
Bytes Rx : 0                 Bytes Tx : 239,212,089
Unicast Rx : 0               Unicast Tx : 330,824
Bcast/Mcast Rx : 0           Bcast/Mcast Tx : 11,569
Errors (Since boot or last clear) :
FCS Rx : 0                   Drops Rx : 0
Alignment Rx : 0             Collisions Tx : 0
Runts Rx : 0                 Late Colln Tx : 0
Giants Rx : 0                Excessive Colln : 0
Total Rx Errors : 0          Deferred Tx : 0
Rates (5 minute weighted average) :
Total Rx (bps) : 0           Total Tx (bps) : 5173088
Unicast Rx (Pkts/sec) : 0    Unicast Tx (Pkts/sec) : 22
B/Mcast Rx (Pkts/sec) : 0    B/Mcast Tx (Pkts/sec) : 1
Utilization Rx : 0 %         Utilization Tx : 00.51 %

 

I think this should not happen in this scale, as a basic principle of switch. This behaviour make us some next problems, but it's a bit off topic now. I can describe it if it can help to solve this problem.

 

It seems that it's somehow connected with the adaptive bonding mechanism on the linux server. Because sometimes this "weird" traffic stops, sometimes is less or bigger. When I let only 2 ethernets from server connected to 1 switch, the problem does not occur.

 

What other information can I provide? I can draw some images, if this is not clear... I can't do much big experiments, because this is production site and we have no spare pieces of this switches available...

 

Thank for your suggestions.