Re: pen vs ldirector

From: Ulric Eriksson <>
Date: Tue, 22 Nov 2005 08:10:59 +0100 (CET)

On Tue, 22 Nov 2005, Mark Jayson R. Alvarez wrote:

> Good day!
> We are currently using ldirectord plus heartbeat, However I would like to try
> another solution and I am looking into pen plus ucarp.
> Questions: 1. I'm not that much familiar with our old lvs setup
> but doing a netstat on those ldirectors, I see no connection
> going to the real servers. But in pen setup, netstat shows that
> the pen balancer is the one who is actually making the
> connection to the real server(it somehow acts as proxy). Now,
> the clients network connection using netstat is only with pen.
> Someone said that the ldirector doesn't perform the actual
> request of the client, it only tells the client on which real
> server it should forward its request(thus the netstat shows no
> output). Am I missing something here? Does this mean all the
> traffic will pass through the pen machine?

Yes. You can think of it as a proxy.

> Another thing, how does pen weighs its connection to determine
> which server is loaded and which is not? The ldirector seems
> like its not based on per connection counts, it only shows all
> of the real servers have equal weights.

From man pen:

        -W Use weight for server selection.

> Lastly, we are going to use pen for squid proxy server
> balancing on port 3128, does pen knows how to check the health
> of this service?

Pen only knows about tcp. It will try to connect to port 3128. If
it fails, the server will be blacklisted for a number of seconds,
30 by default.

        -b sec Servers that do not respond are blacklisted, i.e.
               excluded from the server selection algorithm, for
               the specified number of seconds (default 30).

Since you want to balance two services, apache and squid, you'll
run two pen instances. They will blacklist servers independently.

You can do more sophisticated tests as well. Here is an example:

Received on Tue Nov 22 2005 - 08:10:59 CET

This archive was generated by hypermail 2.2.0 : Tue Nov 22 2005 - 08:11:02 CET