|
On Wed, Feb 04, 2004 at 04:20:20PM -0500, John Reuning wrote:
> I just did some testing on the connection reporting behavior and
> discovered something interesting. This is for an LVS-NAT virtual
> service of http, rr, persistence 300.
>
> The initial page load triggered an ActiveConn count of 1 as reported by
> ipvsadm. The connection status list on the director had SYN_RECV and
> ESTABLISHED for a couple of seconds. After the page finished loading in
> the browser, the connection status switched to SYN_RECV and TIME_WAIT.
> At this point, the ActiveConn count decreased to 0, and the InActConn
> increased to 1.
>
> Here's the weird part: Subsequent page loads from the same browser
> didn't trigger a new active connection. The browser cache was disabled,
> but I even wrote a perl script to repeatedly suck down pages from this
> web server. The InActConn count remained until the SYN_RECV expired on
> the director.
>
> I'm confused by this. Shouldn't new network traffic from a single ip
> address reestablish/reset the connection and be reflected in the active
> connection count?
Yes. Are you sure you are counting it right?
--
Horms
|
