A lot of drops?

[RonniDK]

Hi,

We have a hosting center, with 14 switches, 2 firewalls and a lot Linux-servers.

The network-topology is as follows:

3S-X-S3
| |
3S S3
| |
3S S3
| |
FW-2S----S2-FW <-- example from this
| |
3S S2
| |
2S-X-S2
| |
2S-X-S2

S2 is ProCurve 2900-48GL and S3 is ProCurve 3400cl-48G - FW is Linux-based.

I know it's not optimal in any way, and I have plans for redesigning it.

We have a lot of drops on the links to the Linux-servsers on all 2900-switches (S2) and none to very little on 3400-switches (S3).

The links from the switches to the servers is tagged in one VLAN, and untagged in another VLAN.

Shown below is some of the "show interfaces" from the switch marked in the drawing above.

Status and Counters - Port Counters

Flow Bcast
Port Total Bytes Total Frames Errors Rx Drops Tx Ctrl Limit
------- -------------- -------------- ------------ ------------ ----- ------
1 2,457,943 22,956 0 0 off 0
2 2,612,078 23,352 0 0 off 0
3-Trk2 964,000,111 671,446,478 0 29,248 off 0
4-Trk2 395,611,129 629,918,495 0 10,779 off 0
5 584,234,243 16,714,307 0 71,472 off 0
6 294,874,383 5,369,000 0 71,605 off 0
7 746,758,440 16,582,272 0 71,485 off 0
8 1,049,101,455 1,640,008 0 71,529 off 0
9 242,848,451 22,357,852 0 71,791 off 0
10 3,963,810,031 20,937,502 0 71,634 off 0
11 3,608,809,115 17,627,495 0 71,602 off 0
12 290,359,104 19,827,265 0 71,568 off 0
13 4,151,141,620 29,651,428 0 72,892 off 0
14 3,781,811,466 21,465,250 0 71,813 off 0
15 959,178,384 1,561,576 0 71,598 off 0
16 2,365,153,188 20,415,693 0 71,507 off 0
17 1,739,671,248 30,083,526 0 74,190 off 0
18 3,831,328,429 22,435,041 0 72,096 off 0



We have updated to the newest firmware on every switch in the network, which did not remove the problem, but helped a little.

Does anyone have a clue on this?