|
Hello,
On Wed, 15 Mar 2000, Wayne wrote:
> Has anyone in this mailing list run tests comparable to the
> tests Suresh using? If so, does anyone see this kind of
> problem? We had tested LVS with 400 new connections/s for
> two weeks and not have this problem.
There is no problem with LVS 0.9.7. Used memory near 100%
is normal. The free memory is in "buffers" and in "cached" too.
Anyone can try to create small program that reports how
many MB can be allocated. Try with small values first without
touching the swap. This is simple test that can clear the problem.
>
> At 04:26 PM 3/15/00 -0800, Suresh wrote:
> >There is nothing else running on this machine apart from LVS and ssh.
> >
> >Suresh
> >
> >
> >On Wed, 15 Mar 2000, Wayne wrote:
> >
> > > This is probably due to some other process running on your
> > > box that has memory leak. LVS kernel itself would not have
> > > that kind problem. What kind of processes do you have on the
> > > system?
> > >
> > > At 01:13 PM 3/15/00 -0800, Suresh wrote:
> > > >I have LVS 0.9.7 in use and have some questions re: memory
> > > >usage.
> > > >
> > > >For the last 3 days, it is using all of the available 256M of memory. I
> > > >looked for a correlation with the number of active/inactive connections,
> > > >and couldn't find one. Is there some thing else I am missing?
> > > >
> > > >Our peak active connections are about 1000 on 4 real web servers, and off
> > > >peak about 500 on each web server. I would expect IPVS memory usage to
> > > >vary by # of connections.
> > > >
> > > >The lvs was compiled with 2^16 connections, and runs in VS-DR mode.
> > > >
> > > >Thanks for any pointers.
Regards
--
Julian Anastasov <uli@xxxxxxxxxxxxxxxxxxxxxx>
|
