When you say LARGE HTTP GETs have a problem what is your definition of
LARGE? Are you seeing problems with all files > MTU? Maybe the problem
is being caused by packet fragmentation?
Just a though.
On Wed, 2003-02-12 at 12:58, Joseph Mack wrote:
> Jacob Coby wrote:
> >
> > > LVS requires that all realservers be identical.
> > > This usually means that the realservers are ro.
> >
> > eh? The two realservers are identical,
>
> I meant the content of the service being offered, eg the html files
> on the realservers.
>
> > HTTP, like it says in the subject ;).
>
> OK I have heard of it then :-)
>
> I'm not familiar will uploading with http though.
>
> > > a client is going to change anything on a realserver,
> > > you have to make sure that it's done in one connection
> > > (you probably will have to set persistence) and you
> > > have to be aware of problems that will arise if the
> > > updates are not visible to all realservers.
> >
> > Persistence is 360 sec (6 min), the upload should take no more than 3 sec
> > (this is on a 100mbit Base 100-T ethernet). The file that is being uploaded
> > will be swizzled on the server and put in the database for both real servers
> > to serve up.
>
> OK
>
> > Its not just image uploads that cause the server to hang, large HTTP GETs
> > are also problematic.
>
> I expect it will be easier to test if this is fixed and you should try this
> first.
>
> I don't see anything odd in your tcpdumps.
>
> Joe
--
Matthew S. Crocker
Crocker Communications, Inc. / Vice President
PO BOX 710
Greenfield, MA 01302-0710
Voice: 413-746-2760
Fax: 413-746-3704
Web: http://www.crocker.com
E-mail: matthew@xxxxxxxxxxx
GPG Public Key: www.keyserver.net
|