|
Hi Ken,
thanks for that hint!
With lvs-kiss the solution is pretty easy - so even lady luck worked in
favour ;-)
Within each real-server-definition there are some optional directives.
So the two needed directives are:
RunOnFailure "/sbin/ipvsadm -d -t SIP:80 -r RIP" (for both real servers)
RunOnRecovery "/sbin/ipvsadm -a -t SIP:80 -r RIP" (same)
This does what is needed:
After realizing that a server is down it sets its weight to zero and then
issues that command resp. when it is "up" again vice versa.
Kind regards, Nils
P.S.
Sorry - this e-mail-client destroys thread-information...
> -----Ursprüngliche Nachricht-----
> Von: Ken Brownfield [mailto:krb@xxxxxxxxxxx]
> Gesendet: Dienstag, 13. Juni 2006 18:03
> An: LinuxVirtualServer.org users mailing list.
> Betreff: Re: How can I flush the connection table?
>
> In my experience, I've had to remove the real server entirely
> when I have persistence active. Unless of course you're
> willing to wait an hour :). But I also have had new
> connections continue to trickle in to a zero-weight real
> server in the absence of persistence, so I think there's also
> a general issue where a zero weight isn't literally zero.
>
> I remove real servers within health monitoring in the general
> case, as opposed to zero-weighting them. If zero worked as
> expected, that slightly drastic measure (hard connection
> interruptions, orphans) wouldn't be necessary.
>
> Good luck,
> --
> Ken.
>
>
> On Jun 13, 2006, at 6:13 AM, Hildebrand, Nils, 122 wrote:
>
> > Hi,
> >
> > I subscribed to the mailing list, because I had a major problem
> > with ipvsadm
> > today.
> >
> > I am using ipvsadm v1.24 on a SuSE 9.3 professional (Kernel
> > 2.6.11.4-21.11-smp).
> >
> > For configuring lvs I use lvs-kiss.
> >
> > I am doing a load-balancing for two webservers.
> >
> > Now the problem:
> > The service is configured with a persistency of one hour.
> >
> > After some clients connected through lvs to Webserver #2 the
> > Webserver #2
> > went down.
> > Some seconds later lvs-kiss realized the server is down and sets
> > the weight
> > for that server to zero.
> >
> > BUT: Clients were still trying to connect to Webserver #2.
> >
> > After I looked at "ipvsadm -Lc" I found that their old connection
> > was still
> > there and the timeout-value for Webserver#2 got reset to one hour
> > again
> > after every try.
> >
> > I would have expected that all Webserver #2-connections will get
> > invalidated
> > and that the client-retry would then go to Webserver #1.
> >
> > What's wrong here? [...]
|
