[c-nsp] load-sharing round robin time?

[Dan Letkeman]

I have tried enabling per-packet load balancing, but if I do that then
no pages come up in the browser.  So I did a tcp-mss adjust on the
interface and still no difference.

topology:

lan----squid box----2621 router-------4 827 modem's(nat & adsl)


Dan.

On Thu, Sep 11, 2008 at 9:12 PM, David Coulson <david at davidcoulson.net> wrote:
> You can set it to use per-packet load balancing instead, assuming all of the
> paths are essentially the same (otherwise you get out of order packets,
> which may not be what you want).
>
> Is the squid box on the 192.168.11.x subnet? If you have ip redirects
> enabled, then the squid box will actually route directly to one of the
> gateways, rather than through the 2621... Not sure how your environment is
> build - Maybe a routing table and some other interface configs would help?
>
> Dan Letkeman wrote:
>>
>> Hello,
>>
>> I'm doing load-sharing on a 2621 router with ios 12.3(26).
>>
>> ip route 0.0.0.0 0.0.0.0 192.168.11.251
>> ip route 0.0.0.0 0.0.0.0 192.168.11.252
>> ip route 0.0.0.0 0.0.0.0 192.168.11.253
>>
>> This was working just fine, but now we implemented a squid cache just
>> behind the router and it strips the source ip, so all of the requests
>> through the router all look like they are coming from the squid box
>> now.  What is happening now is the squid box is randomly switching
>> from route to route, but it's taking about 10 minutes to switch from
>> each route.  So watching the graphs on the three routers and its only
>> really using one route at a time. Is there a way to change the time
>> limit for switching routes to make it switch faster?
>>
>> Thanks,
>> Dan.
>> _______________________________________________
>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>
>