[squid-users] Tr : Re: Loadbalancing multiple SQUID parents using a parent persistence selection per client connection

----[ Just a forward to keep historic on squid mailing list ]---

Hi ratz,

>> +----------------------------+
>> | Internet |
>> +----------------------------+
>> | | |
>> +----------------+ | +-----------+
>> | | |
>>+----------------+ +----------------+ +----------------+
>>| PROXY-PARENT-1 | | PROXY-PARENT-2 | | PROXY-PARENT-3 |
>>+----------------+ +----------------+ +----------------+
>> ...................................................
>> ...................................................
>> ...................................................
>>+---------------+ +---------------+ +---------------+
>>| SQUID-PROXY-1 | | SQUID-PROXY-2 | | SQUID-PROXY-3 |
>>+---------------+ +---------------+ +---------------+
>> | | |
>> +--------------+ | +---------+
>> | | |
>> +----------------------------+
>> | LoadBalancer |
>> +----------[ VIP ]-----------+
>
>>* 3 Antivirus PROXY-PARENT parents of 1.
> ^^^^^
> of 1 what?
>
>Do I understand if I say that every SP can connect to each of the three
PP?

Yes they can but SP must be sticked to a PP during a specific client
session (during the browse over a website).

Clients (webbrowsers) are parametered with connect the LVS VIP/port (for
instance 8080). LVS garanty, using persistence, that all traffic is sticked
to a specific SP. But I need to be able to garanty the same things on the
SQUID parent selection. That a client (represented by its IP address),
still sticked to a specific PP. That way for a specific client, runing this
topology I preserve & keep the SAME PROXY path during a client session.

=> My PP are installed in a DMZ with public routed over Internet IP
addresses. So if the IP address for a specific client change (on the remote
server connected eg www.linux-vs.org for example) during a client session
its visible IP address to the Internet will change during its session =>
The fact is that many Internet web sites doesn't like that kind of live
change (AOL mega-proxy are working like this, this is why so many users
have problems on some sites).

Does it sound more comprehensive ? because my too poor english level make
me some fault :)

>>My question is located on the SQUID parent definition. I want that each
>>SQUID loadbalance incoming traffic to parent pool. My setup must work
with
>
>Incoming from where? Client-side?

Yes => LVS loadbalance Layer3 so SP selected by LVS will see the client IP
address.

>>I would like to know if it is possible to make a connection persistent
from
>>a specific client to a specific parent ? Is it possible to stick traffic
>>from a client to the same parent (without round-robin algo replay for all
>>GET query comming from the same client) ?
>
>Going over the SP(i) first? What does the network scheme look like
>(IP-addressing)?

Yes web browsers use LVS SQUID VIP/VPORT. LVS stick client to a specific SP
using persistence.

=> I just want to know if when using SQUID round-robin parent declaration,
SQUID doesn't replay the rr algo when the stream are coming from the same
IP address (web browser).

I am not really sure that SQUID act on that point as I suspect (wish ) :/

Thanks a lot if you have any inputs on that point :)

Best regs,
Alexandre

_______________________________________________
LinuxVirtualServer.org mailing list - lvs-users@LinuxVirtualServer.org
Send requests to lvs-users-request@LinuxVirtualServer.org
or go to http://www.in-addr.de/mailman/listinfo/lvs-users
Received on Tue Mar 19 2002 - 05:45:40 MST