[f-nsp] Latency issue

[Eric Roach]

Vladimir,

The paste from a Force10 VRRP problem I had in the past is not related entirely to your post, but it did cause me some headaches. The issue was pinging the router interface and I was getting 15 ms delay. I wonder if the code on both routers is the same? Here is the Force10 issue.

Symptom: Packet Loss When Pinging the Virtual IP Address
When pinging the virtual IP address of a VRRP group, packet loss is experienced. 
Troubleshooting Steps
Force10 does not support 100% pings to VRRP virtual addresses. The reasons are:
	* Pings to the virtual IP address are forwarded to the CPU RP2 on the Route Processor Module (RPM) using the network entry in the line card CAM. By design, such RP2-bound packets are rate-limited to protect the CPU from DOS attacks and other unwanted packets. 
	* The E-Series supports hardware and software throttling to limit packets destined to any of the RPM CPUs -- CP, RP1, or RP2. Such throttling ensures that one packet type does not consume an excessive number of CPU cycles. The throttling mechanism works by placing packets in one of several predetermined queues, which have inherent rate limits.
	* Eric Roach




________________________________
From: Vladimir Dyachenko <vlad.dyachenko at gmail.com>
To: foundry-nsp at puck.nether.net
Sent: Wednesday, June 10, 2009 2:49:01 PM
Subject: [f-nsp] Latency issue

Hello,

I got a strange problem for some time now.

I have two RX-4 routers, each one has an uplink to an upstream and get a full BGP table. Both are connected together thru BGP/OSPF and run VRRP owner/backup.

Then I have a switch connected to both routers and my servers are connected to that switch.

-----------------       -----------------
|               |       |               |
|     BR 1      |-------|      BR 2     |
|               |       |               |
-----------------       -----------------
               \         /
            -----------------
            |               |
            |      SW 1     |
            |               |
            -----------------

From Internet when I connected from BR1 (with upstream 1) everything works fine, latency is normal. But when I connect thru BR2 (with path over upstream 2) my latency get 80ms for now reason on the last hope (between the provider interface on my side and the target).

I run a VLAN '2',BR1 has spanning-tree priority 200 and BR2 has spanning-tree priority 100 and my switch has spanning-tree priority 0.

Note : this only happends for upstream 2 on BR2, and when I turn off router 1 (and thus disconnected upstream 1) latency is again normal for upstream 2.

Any idea ? 

Best regards.

Vladimir
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/foundry-nsp/attachments/20090610/457adc68/attachment.html>