On Wed, 21 Feb 2007, Matthew wrote:
is the problem seen from the same client machine during this time or are
random people having the problem
It seems almost random. Here in our office, we have one DSL
connection and all our desktop computers experience this behavior. From my
house, Cable, no issues. From a friend in another city, no issues. From a
customer in Chicago, issues present. ping's and traceroutes to/from all
machines (including DIR and RS) are super fast.
OK so the problem depends on the client's location (I assume
they're all doing the same thing as client).
(Since it's only occuring with some clients, I'll assume the problem is not
in LVS, but somewhere else. I know there are flaws in this assumption, but
it's somewhere to start.)
But it's gotta be LVS/DIR right? Since accessing the same page via
DIR vs RS gives differences, and the DIR is the point of entry, wouldn't that
make it a LVS/DIR problem?
It's an interaction between the setups of certain customers
and LVS. We don't know whether it's LVS or the customer's
setup yet. To show that it's LVS we have to show that the
customer's hardware is behaving properly. There are plenty
of precedents that it doesn't. If the customer's hardware
isn't working properly, it will be relatively easy to fix.
Because of the triangulation involved in LVS-DR/Tun, the tcp
agreements have to be maintained more tightly. In this case
a bare server may not have problems when and LVS will. I'm
only looking at the non-LVS part of your setup for a
solution, because it's going to be a lot quicker to
fix/exclude than looking in the LVS.
Can you do exactly the call from home (or good site) that your customer in
Chicago has problems with?
Yep. We are all doing the same call, that speedtest.html page. Feel
free to try it yourself.
I can't see the traffic on the realservers from here.
If you can't reproduce the problem from home, you'll have to get the
Chicago person on the phone and tcpdump the connection.
The person in Chicago is a pretty important customer and I don't want
them to know we are having issues. Since our local office is having problems,
I'll tcpdump an office d/l and then one from home. Will post when I get them.
Otherwise you may have weird hardware in the connection (I know this won't
be easy to track down)
If the hardware was the problem and the hardware hasn't changed since
day 1, wouldn't this problem have affected us from day 1? This problem has
only happened once in the past; about a week ago. It affected us for about
2-3 days then seemed to clear itself up. Now its happening again.
well something is different; hardware, software...
Joseph Mack NA3T EME(B,D), FM05lw North Carolina
jmack (at) wm7d (dot) net - azimuthal equidistant map
generator at http://www.wm7d.net/azproj.shtml
Homepage http://www.austintek.com/ It's GNU/Linux!