|
Ok. I'm not sure exactly how to respond to this, but I'll try. Pretty
much any web server running new applications will "require" persistence.
I'm actually not a unix admin by trade, I'm a php programmer, so this is
one area where I actually know what I'm talking about. Not everything on
a web site uses server side sessions, but some things do. Some things
need to. In fact, you'd have to be running a very basic static-content
website to not need to use server-side sessions and cookies. I remember
back in the day when I used to "log in" people to my website by storing
their IP in a database. I remember how many problems that caused because
of NAT in other people's offices. Now that I can use php server-side
sessions, that is no longer a problem. So I guess my point is that by
saying that persistence is unnecessary, you are also saying that
sessions are unnecessary, which is not true.
So why try to load balance when your web applications are using
server-side sessions? My answer is the same reason I wanted to load
balance in the first place. Having sessions certainly doesn't negate the
need for a load balancer. In fact, if you are using heavy server-side
scripting in your websites enough to use a lot of sessions, then you
need load balancing MORE than most people.
Sticking with simply IP based user management is like saying "We have a
great application on the internet for everyone to use, but only one
person in every office can use it, because if two people try to use it
then the web server still sees them as one person and it causes
problems."
So is persistence a marketing gimmick? I think not. I'd say it is more a
SOLUTION to a NEED. However, lucky for me, it is not necessary beyond
ip-based persistence. Because like Jeff said, it's not that big of a
deal since it just means that everyone from that NAT'd office will go to
one webserver until they are all done and next time it switches them to
the other server. It will still be load balancing but just not as
perfectly. And I can deal with the server-side sessions by using rsync
just like I'll deal with every other aspect of running two servers with
the same content. That + IP based persistence since rync wouldn't be
fast enough to handle someone going from page to page making changes to
their session and hitting a different server each time. Anyway, I should
be ok when using server-side sessions + LVS persistence. I think.
- Doug
-----Original Message-----
From: lvs-users-admin@xxxxxxxxxxxxxxxxxxxxxx
[mailto:lvs-users-admin@xxxxxxxxxxxxxxxxxxxxxx] On Behalf Of Malcolm
Turnbull
Sent: Monday, August 12, 2002 10:36 AM
To: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Subject: Persistence is a silly marketing gimick for everything apart
from SSL.. TROLL
TROLL----
SSL likes persistence so that it doesn't need to re-key for each
connection...
Nothing else needs it.
If you web server requires persistence (i.e. local sessions) then why on
earth bother with a load balancer.
END OF TROLL----
Jeff Warnica wrote:
| Im not at all an expert, but I dont see how this would be possible
with
| LVS.. LVS works on a much lower level then any of apache/php/ssl
session
| type information. IMNSHO that the director is blistfully unaware at
the
| nature of the traffic its directing, and the real servers (can be)
| blistfully unaware that there in a cluster is a Good Thing.
|
| While Im sure there are situations in various enviroments that will
| always cause an unbalance (ie, on an intranet from a NAT'd office),
the
| situations that I can imagine are somewhat self inflicted. NAT due to
IP
| shortage is something of a hack, and if your (or your corps) hack
| confuses your director then you could consiter putting the director
| inside the NAT'd area.
|
| For public (ie, internet) servers, I cant see NAT induced inbalances
| substaining over a long time. (ie, no longer then some time related to
| the average web visit - time enough for eyeballs to leave your site
and
| new eyeballs to be sent to a different RS).
|
| But I might have no idea what Im talking about.
|
| On Mon, 2002-08-12 at 12:14, Doug Schasteen wrote:
|
|>One of the features that I saw a lot when looking at commercial load
|>balancers was the ability to use persistence based on cookies, session
|>IDs, and ssl IDs instead of just IP addresses. The reason why is best
|>explained right here. If a lot of people work in an office that has
only
|>one public IP address, then you get lots of load imbalance. Is it
|>possible for LVS to do persistence based on apache/php sessions? Or
|>based on SSL sessions, or anything like that? Of course, I don't know
|>how these types of persistence would help when balancing an e-mail
|>server.
|>
|>
|
|
|
| _______________________________________________
| LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
| Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
| or go to http://www.in-addr.de/mailman/listinfo/lvs-users
--
Regards,
Malcolm Turnbull
IT Manager
Crocus.co.uk Ltd
01344 629661
07715 770523
http://www.crocus.co.uk/
_______________________________________________
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://www.in-addr.de/mailman/listinfo/lvs-users
|
