Re: super slow speeds from director

To: " users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: super slow speeds from director
From: Roberto Nibali <ratz@xxxxxxxxxxxx>
Date: Thu, 22 Feb 2007 22:04:09 +0100
Hello Matthew,

What OS are the clients running?

    Doesn't seem to matter. I'm Mac OSX at home, WinXP at work.

But only fetching from the office you've mentioned it to be slow.

No handshake visible, please capture the whole TCP flow.

    How? What command did I not issue to tcpdump?

No TCP/SYN visible in this specific dump compared to the others. You need to start tcpdump before you fetch the page, like you did for the other 3 dumps. Since this is the reportedly slow flow, so I'm also interested in the connection setup phase and in the times.

Hmm, could you add at least one '-v'? To me it's not enough information to figure out what's going.

    Will repeat and add -v.

Hopefully we get more information.

Now this is interesting! What strikes me as odd in your dumps is the client address with a trailing '0', aka

Yea, that's our office connection with SBC AT&T. When I visit that's what gets reported.

Could I try from somewhere? Is there an URL to test fetch?

change your IP address of the test client

    I'd like to, but SBC has us on DHCP so I can't dictate my IP.

Maybe you can issue a ticket to them to not give you such IP addresses?

Also note, that testing
from within the cluster does not give you proper results due to the asymmetric routing.

Not sure what you mean by that. The cluster is in Dallas and I did client testing from Houston and San Antonio. If you are referring to something else, please clarify.

Sorry, I missed that you're using LVS_TUN. Please ignore my comment above.

Roberto Nibali, ratz
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc

<Prev in Thread] Current Thread [Next in Thread>