GE redundancy through fibre for box-to-box redundancy

[dengqi]

According to Cisco documentation, you need to use crossover cable to connect master and redundant CSS for box-to-box redundancy. May I know whether the fibre can be used instead of crossover cable? And how about the ISC link for ASR? Can I use fibre also?

[Gilles Dufour]

ASR is not supported with box-to-box redundancy.

It was designed for Vip redundancy.

You can use fiber for the ASR link.

For box-to-box redundancy you can know use fiber with the latest version - but it is useless since this link is only used for sending hearbeat - very low BW.

Regards,

Gilles.

[dengqi]

the reason that I want to use fibre is that the two CSSs will be placed in two different buildings a few hundred meters away. So I'm think use SM fibre to link these two CSS as the redundancy link for box-to-box redundancy.

and for redundancy-l4-stateless, does it means a stateful failover for TCP session and user does not need to reestablish connection to server?

Thanks

[Gilles Dufour]

with box-to-box redundancy this is indeed stateless which means client needs to reconnect to the server after the css failover.

If you need statefull failover, use vip redundancy with ASR.

It requires more commands, but it gives you faster failover and stateful failover.

Gilles.

[dengqi]

Gilles,

CSS Redundancy Configuration Guide (Software Version 7.30) said that:

Use the redundancy-l4-stateless command in content or group configuration mode to enable stateless redundancy failover in a box-to-box redundancy or a VIP and virtual interface redundancy configuration. Stateless redundancy failover allows critical TCP/IP traffic to continue in case of a failure at the load-balancing CSS by allowing the backup CSS to set up a mid-stream TCP flow. This feature is disabled by default.

The default behavior of a CSS is to set up load-balanced TCP flows only when it receives a TCP frame that begins with SYN. When stateless redundancy failover is enabled and a failover occurs, the backup CSS establishes a mid-stream flow for any existing TCP sessions. The CSS still exhibits the default behavior for all new flows. To restore the default behavior of the CSS for all flows after issuing the redundancy-l4-stateless command, use the no redundancy-l4-stateless command.

Does this mean that client does not need to reestablish connection to server after failover?

Thanks

Deng Qi

[Gilles Dufour]

Deng,

yes you can use this command to tell the CSS not to drop mid flow packets after a failover.

The CSS will attempt to recreate a flow.

However, this is not always possible.

If the CSS receive a packet from client X to vip V.

How does it know it should be sent to server Y ?

There is no way. It will just send to one server and if you're lucky this is server Y. If not, the newley selected server will reject the packet because not part of any connections to/from this server.

As you can see, you get very limited benefit with this command.

ASR is the only solution to provide statefull redundancy.

Regards,

Gilles.

[dengqi]

Hi Gilles,

Thanks for your reply. I have much clearer understanding now. If I use arrow-point cookie as the sticky method, will the new CSS understand the cookie sent by client and forward client connection to the right server?

Thanks

Deng Qi

[Gilles Dufour]

The arrowpoint cookie contains the server ip address or the server string.

So, yes, the backup CSS should be able to use the cookie to forward the traffic to the right server.

Gilles.